A few more podcasts I keep returning to. Not really sleeping podcasts, though they often accompany the end of the day. More like little lanterns for the mind before it gives up and becomes a vegetable.

The podcast I listen to most these days is probably not about business, politics, technology, or even history.

It’s about daughters.

Recently I listened to an episode of Conan O’Brien Needs a Friend where Conan and Matt Gourley spent the first fifteen minutes talking about their daughters.

Then an actress showed up and they did an interview. The interview was fine. The daughters were the story.

The older I get, the more I notice that the most interesting people eventually end up talking about the same things. Family. Friendship. Regret. Gratitude. The weird terror of loving someone so much that they become a permanent vulnerability.

Conan is a delight because he spent decades becoming one of the funniest men in America and now often sounds like somebody’s slightly bewildered dad.

Which brings me to the greatest podcast ever made: the BBC’s In Our Time.

For years, especially during Covid, it was my church.

Every episode is the same. Melvyn Bragg gathers three academics in a room and asks them about some obscure topic that normal people are supposedly not interested in.

The Hanseatic League. The Abbasids. Entropy. The Bronze Age Collapse. The history of zero.

Nobody interrupts. Nobody is building a personal brand. Nobody says “actually.”

It is a place without morons.

Covid was strange because I was simultaneously spending my days with two small daughters and my evenings listening to medievalists explain fourteenth-century trade routes. Looking back, both activities had roughly the same psychological effect. They reminded me there was a world larger than whatever panic happened to be on television.

If In Our Time is my church, then Bob Dylan: Album by Album is my monastery.

There are people who know every Bob Dylan record. Then there are people who know every Bob Dylan bootleg. Then there are the people who make Bob Dylan podcasts.

These people have crossed beyond fandom into something approaching religious vocation. I admire them enormously.

On the Media occupies a different place.

Every week it reminds me that the story is often not the story.

The thing that happened is interesting. The way people talk about the thing that happened is usually more interesting.

The machinery behind public understanding is one of the most important subjects in the world, and one of the least discussed.

Then there is The Rest Is History. Of course.

History is one of humanity’s great inventions. It allows us to discover that everything happening today happened before, but with worse dentistry.

The hosts somehow manage to combine scholarship, enthusiasm, and the energy of two men trapped at a dinner party who have just discovered nobody is making them leave.

Acquired is the business equivalent.

Most business journalism is about quarters. Acquired is about decades.

Most people underestimate how much more interesting a company becomes once you tell its story as a civilization rather than a stock ticker.

I have learned more from Acquired’s deep dives than from many business books.

Song Exploder remains one of the most elegant ideas in podcasting.

Take a song. Ask the people who made it how they made it. That’s the whole show.

The lesson is larger than music. Almost everything interesting becomes more interesting when you learn how it was built.

The same is true for comedy. Good One is a podcast where comedians explain jokes.

The Zach Galifianakis episode is particularly wonderful because he reveals something many people miss: comedy is often less about wit than about architecture.

A joke is a machine. The comedian is an engineer.

Unclear and Present Danger is another favorite.

The premise is simple: revisit the paranoid political thrillers of the late twentieth century and ask what they were afraid of.

The answer is often that America keeps inventing new anxieties while preserving the old ones in amber.

Finally, Jacobin, and specifically Jacobin Radio.

I don’t always agree. That is partly the point.

A healthy intellectual diet requires exposure to people who think differently from you, provided they are intelligent, serious, and arguing in good faith.

The internet has convinced us that disagreement is dangerous. In reality, disagreement is often how thinking happens.

Looking at this list, I notice that almost none of these podcasts are really about their stated subjects.

History podcasts are about human nature. Business podcasts are about ambition. Music podcasts are about creativity. Media podcasts are about reality. Comedy podcasts are about structure. Political podcasts are about disagreement. And podcasts about daughters are about love.

Which, in the end, turns out to be what most things are about.

— The Bugle for understanding stupid Trump people.

— My podcast “In the Know” for understanding stupid business people.

The Bugle is what happens when several extremely educated British people decide the only sane response to modern politics is sustained civilizational mockery. It’s less “news analysis” and more “Rome collapsing but everyone has microphones.” You come away understanding that half of politics is theater, a quarter is grievance, and the remaining quarter is men named Doug posting eagle memes next to pickup trucks.

Meanwhile, “In the Know” explores the equally bizarre anthropology of business civilization. Venture capitalists. Founders. AI prophets. Wellness futurists. Productivity shamans. Men who say “let’s jam on this” while slowly becoming PowerPoint. It’s nominally about innovation, but actually about ambition, status rituals, optimism, self-deception, and the quasi-religious belief that apps can somehow redeem mankind if the seed round closes in time.

One podcast explains why democracy feels like a casino attached to a Facebook comments section.

The other explains why every startup pitch now sounds like a TED Talk delivered during a ketamine-assisted ayahuasca retreat at Soho House.

Together they form a complete liberal arts education for surviving the 2020s.

Also:

— The Rewatchables for understanding why men can explain the 1995 movie Heat with the seriousness of medieval theologians debating the nature of God.

The Rewatchables is secretly not about movies at all. It’s about memory. Male friendship. Cable television. The anthropology of quoting scenes from Jaws for thirty years. Bill Simmons and the rotating cast approach films with the energy of monks preserving sacred texts after the collapse of civilization. Entire emotional universes are reconstructed from seemingly meaningless details like “that guy actors,” “apex mountain,” or whether the movie would somehow improve if Wayne Jenkins from The Wire burst into the scene screaming profanities.

It’s one of the last surviving corners of culture where people are still allowed to love things without embarrassment or irony.

Movies as oral tradition. Cinema as folk religion. Dad lore with a soundboard.

One of the great poems, by Baudelaire, to remember when you are preposterous and forlorn

TO A WOMAN PASSING BY

The boom and yelping in the street had deafened me.

Majestic in her grief and widow’s weeds, so grand,

So slim, the woman passed, one ostentatious hand

Lifting festoon and hem to sway them casually,

Lightfooted, svelte, her ankle nonpareil.

Me? I was drinking up, preposterous, forlorn,

From her grey eyes, like clouds where hurricanes are born,

The sweetness that ensnares, the pleasure that can kill.

One flash… then night—O beauty, must you pass from me?

One look and I’m reborn; my life is changed forever;

Shall I see you again in all eternity?

Somewhere very far from here! too late! Or never!

You know not who I am; I know not where you go.

O you, I could have loved you. And I know you know.

But don’t forget why I love the sadhu on the street

It could be any of us

And if we all believed the high moral aims of poverty, devotion, purity, charity, and freedom… it would be all of us

A half-naked fakir, said Winston Churchill about Mahatma Gandhi

Gallipoli: Feb 1915–Jan 1916 overall; the land campaign began 25 Apr 1915 and ended 9 Jan 1916. Dead included Allied troops from Britain, Ireland, Australia, New Zealand, India, Newfoundland, France, and Ottoman/Turkish forces. Roughly 8,000+ Australians, 2,700+ New Zealanders, many thousands of British/French/Indian troops, and huge Ottoman losses; total deaths were well over 100,000 across both sides. Churchill was First Lord of the Admiralty and became politically tainted by the disaster.  

Bengal famine: 1943, in wartime British India. Usually estimated at about 3 million deaths from starvation and disease. Churchill’s defenders dispute sole blame; critics argue British wartime policy, shipping priorities, rice denial policies, inflation, and imperial indifference worsened it gravely.  

UK retreat from France: the key event is Dunkirk / Operation Dynamo, 26 May–4 June 1940. About 338,000 Allied troops were evacuated, including about 198,000 British and 140,000 French and Belgian troops. It was a rescue wrapped around a defeat.  

Chronology knife-edge:

1915–16: Gallipoli catastrophe.
1940: Britain flees France at Dunkirk.
1943: Bengal famine.
1947: Britain leaves India.

Dat is een harde reeks — “That is a harsh sequence.”

The Dutch there is not irrelevant.

Neutrality, the great language of the Swiss and… of Belgian cousins. We are small and weak, let us not take sides (we need money, we like living in London). One country is impossible mountains, never invaded. The other is flat sand. A highway to Holocaust. A philosophical imagination:

---

SOCRATES: Tell me, Leopoldes — what is neutrality?

LEOPOLDES: The right of a small nation to remain outside the quarrels of great powers.

SOCRATES: A legal definition. I asked for a real one.

LEOPOLDES: Then neutrality is survival.

SOCRATES: Ah. Survival without choosing.

LEOPOLDES: Small countries cannot afford grand moral gestures.

SOCRATES: Yet history forces gestures upon them anyway.

Let us begin with Belgium in the first war.

Germany demanded passage through Belgium to invade France.

Belgium refused.

Germany invaded anyway.

Belgian towns burned. Civilians were shot. Libraries destroyed. Refugees poured across Europe.

A small country became a martyr.

This much is true.

LEOPOLDES: Entirely true.

SOCRATES: And yet I wonder whether neutrality itself contained the seed of catastrophe.

Belgium trusted treaties the way a lamb trusts fences.

LEOPOLDES: Treaties are civilization.

SOCRATES: No. Treaties are civilization during peacetime.

War reveals what civilization actually is.

Belgium believed law could stop momentum. But when German generals looked westward, Belgian neutrality became not a sacred principle but merely a road.

LEOPOLDES: What alternative existed? We were small.

SOCRATES: Smallness is not innocence.

Nor is neutrality wisdom.

Belgium stood between Germany and France while insisting it was not part of their struggle. Yet geography itself contradicted this fantasy.

You sat in the doorway and declared yourself uninvolved in the fire.

---

The wind sharpens.

---

SOCRATES: Still — in the first war Belgium fought bravely once invaded. Liège resisted. The Yser held. Belgium became heroic precisely because it was overwhelmed.

But the second war is darker.

There, neutrality was no longer innocence.

It was repetition.

LEOPOLDES: Belgium remembered 1914.

SOCRATES: Trauma often disguises itself as prudence.

After the first war, Belgium drew the wrong conclusion. It decided not that Germany was dangerous, but that entanglement itself was dangerous.

So Belgium drifted back toward neutrality in the 1930s.

You weakened military coordination with France.

You hoped to avoid provoking Germany.

You attempted once again to stand beside history while history accelerated around you.

LEOPOLDES: We built defenses.

SOCRATES: Forts.

Concrete theology.

Europe worshipped fortifications between the wars because fortifications permit the illusion that war will remain orderly.

The Maginot Line.

Eben-Emael.

Lines on maps.

But tanks do not respect geometry any more than ambition respects treaties.

LEOPOLDES: Belgium could not defeat Germany alone.

SOCRATES: Then why pretend isolation increased safety?

That is what fascinates me most about neutrality: it is often fear disguised as dignity.

Belgium feared becoming a battlefield again.

So it delayed deep cooperation with France and Britain.

And what followed?

Germany invaded through Belgium again.

France rushed north according to the old plan.

German armor cut through the Ardennes.

France collapsed with astonishing speed.

Europe burned.

Neutrality did not prevent catastrophe. It merely postponed clarity.

---

LEOPOLDES: You make Belgium sound responsible for France’s fall.

SOCRATES: Not responsible. But contributory.

France committed its own suicide through arrogance, bureaucracy, exhausted demographics, and doctrinal rigidity.

Yet Belgium helped create the conditions for surprise by insisting on strategic ambiguity.

A nation cannot simultaneously occupy the hinge of Europe and claim to be merely decorative.

---

A silence follows.

Far away comes the sound of artillery that no longer exists.

---

SOCRATES: And now we approach the forbidden subject.

Sympathy.

Tell me, Leopoldes: were there Belgians who admired aspects of the Nazi order?

LEOPOLDES: In every occupied country there were collaborators.

SOCRATES: Do not dilute the question.

I ask because neutrality often produces a peculiar moral fog.

When a nation defines itself primarily by avoiding conflict, conviction weakens.

Some Belgians saw Germany not merely as invader but as future.

Authoritarian order.

Anti-communism.

Hierarchy.

Stability.

In Flanders especially, some nationalist movements collaborated enthusiastically.

The rhetoric was familiar across Europe:
Perhaps accommodation is wiser than resistance.
Perhaps Germany cannot be stopped.
Perhaps modernity itself belongs to the strong.

LEOPOLDES: And many Belgians resisted heroically.

SOCRATES: Yes. Humans divide under pressure.

But neutrality before invasion creates strange habits after invasion.

A country trained not to choose often struggles to choose quickly when choice finally becomes unavoidable.

---

LEOPOLDES: You condemn small nations for trying to survive among giants.

SOCRATES: No.

I condemn illusions.

Belgium believed neutrality exempted it from history.

But there are locations on earth where neutrality becomes impossible because geography itself is political.

Belgium was one of those places.

Poland another.

Ukraine now, perhaps.

The plains of Europe are invitations.

---

The philosopher kneels and touches a poppy.

---

SOCRATES: England remembers the poppy as sacrifice.

And rightly so.

But sacrifice alone is a dangerous memory.

A nation that remembers only how it suffered begins to imagine itself morally pure.

Belgium remembers burned towns.

But does it remember the Congo?

LEOPOLDES: Again Congo.

SOCRATES: It must always return.

For while Belgium pleaded for the sanctity of treaties in Europe, it governed millions in Africa without consent.

Belgium demanded sovereignty abroad while denying it elsewhere.

This contradiction matters because it reveals something terrible about modern Europe:

Every empire believed brutality became immoral only when directed inward toward Europeans.

---

LEOPOLDES: You reduce all morality to hypocrisy.

SOCRATES: No.

To tension.

Belgium was both victim and perpetrator.

The British Empire fought Hitler while ruling colonies.

France defended liberty while suppressing Algerians.

Germany produced Beethoven and Auschwitz.

Civilization is not moral consistency.

It is merely complexity with architecture.

---

LEOPOLDES: Then what should a nation do?

If neutrality fails, empire corrupts, alliances entangle, and memory deceives — what remains?

SOCRATES: Honesty.

Not innocence.

Honesty.

A mature nation says:
We were brave.
We were frightened.
We resisted.
Some collaborated.
We suffered.
We exploited others.
We trusted treaties too much.
We prepared too little.
We hoped history would pass around us.
It did not.

That is adulthood.

Most nations prefer mythology because mythology is emotionally ergonomic.

---

The sea grows dark.

---

SOCRATES: Tell me finally, Leopoldes:

What is neutrality?

LEOPOLDES: I no longer know.

SOCRATES: Then perhaps you are beginning to understand it.

Neutrality is not peace.

Neutrality is merely the hope that history chooses another route.

And Belgium learned twice that Europe often travels through the same doorway.

---

The wind scatters poppy petals across the stones like drops of blood.

Neither man speaks again.

An essay from Tokyo, dears. May 2026

Every civilization has poetry, religion, bread, bureaucracy… and a word for “please contain this mess.”

In America, it’s a diaper. In Britain, it’s usually a nappy. In India, you’ll hear both, but “diaper” is common in stores and online shopping. In adult-care retail, everyone suddenly becomes coy and says incontinence pants, briefs, pads, or protective underwear. Language puts a little velvet curtain around the subject. Very British. Very pharmaceutical. Very human.

United States: “Diaper”

The American word is diaper, and it originally referred not to babies, but to a kind of patterned cloth. The modern baby meaning came later. See Merriam-Webster on “diaper” and Oxford English Dictionary on “diaper”.

Adult versions are often marketed as:

• Adult diapers
• Briefs
• Protective underwear
• Incontinence underwear
• Pull-ups

United Kingdom: “Nappy”

In Britain, babies wear nappies. Adults, however, are more likely to be pointed toward incontinence pants or adult nappies. “Adult nappy” is the literal phrase; “incontinence pants” is the shop-safe phrase. Search both.

Dictionary backup: Merriam-Webster defines “nappy” as chiefly British for diaper.

India: “Diaper,” “Nappy,” and Brand-Language

India is gloriously multilingual, so shopping language often follows the platform: “diaper” is extremely common online, while “nappy” may appear in baby-care contexts. For adults, search terms like adult diaper, elderly diaper, incontinence diaper, and brand names such as Friends, TENA, Depend, or iD may work better than delicate euphemism.

Example: iD’s adult diaper launch in India.

Australia, Ireland, and Friends

Australia and Ireland largely follow the British pattern: nappy for babies, adult nappy or incontinence product for adults. The empire left behind railways, tea habits, and apparently a different word for absorbent undergarments. History is weird.

The Great Euphemism Ladder

What to Search on Grocery Apps

If you’re in the UK, search:

• TENA pants — probably the best search
• Incontinence pants
• Adult nappies
• Pull up pants
• Protective underwear

Why the Words Change With Age

For babies, the words are cute: nappy, diaper, pull-up. For toddlers, they become developmental: training pants. For adults, the words become discreet: briefs, pads, continence care. Same basic technology, different social packaging.

That’s the funny thing about English: it does not just name objects. It manages embarrassment, class, region, age, and the aisle you’re standing in at 11:47 p.m. while Deliveroo asks whether you meant “incontinence underwear.”

Useful Links

• Merriam-Webster: Diaper
• Merriam-Webster: Nappy
• Oxford English Dictionary: Nappy
• Adult diaper overview
• Wikimedia Commons: Baby diaper images

Final wisdom: in America, say “adult diapers.” In the UK, search “incontinence pants.” In India, search “adult diapers.” In all countries, search by brand if you want to avoid linguistic slapstick. TENA is the Esperanto of this little world.

Don’t worry, this post isn’t about you https://apple.news/AYOhz_3t9QYGKXY_KqNnJtA https://archive.md/2026.05.19-144242/https://www.theatlantic.com/magazine/2026/06/conservative-masculinism-misogyny/686939/

My friend and I recorded a long time ago, one song. And here it is released.

“Oops, America starts a war on Iran and loses. Hormuz is locked, the US ‘retreats’ and it’s declared a victory.”

But everyone will know that the US can no longer do what it wants.

Flashback: Suez Crisis and the UK. End of the Pound. Deleveraging. Decline of the UK.

Today: short term jump as world destabilizes (flight to Treasuries) then slow bleed end of the Dollar. US assets decline relatively. Scary!

Why:

war / instability → inflation pressure

→ higher rates

→ worse fiscal math

→ more supply of Treasuries

→ marginal buyers hesitate

Nah, Claude wrote me this script.

imessage_cleanup.sh.JPG  (download it, unzip it, rename it to remove the .JPG so it’s just imessage_cleanup.sh)

Put it on your OSX and then type:
chmod +x imessage_cleanup.sh
sudo bash imessage_cleanup.sh
Or write your own this way:
“write me a super detailed script I can run on my osx as root sudo to clean up my imessage storage which is using 20GB of icloud. I want to detach all images and put them in a folder on my desktop organized by the sender (guess the sender based on conversation and my contacts db) then I want to delete all the images from the icloud/imessage collection and just have them in my desktop folder so I can back them up in a different way. the script should be extremely robust with. onscreeen terminal UI, options, selcgtors, error handling, logging, and safety before irreversible actions.”

I’ve been writing online since 1998. I’ve built companies, invested in others, taught, experimented, and tried to write down what I was learning along the way. Many of the questions I get are already answered somewhere below.

This page is a comprehensive source guide. If you read a few of these first, your email will be sharper, more contextual, and much easier for me to respond to well.

Start Here

• About – background, companies, teaching, and context.
• Cold Email to Me — Start Here
• Main Weblog
• Everything I’ve written (author archive)

Books & PDFs

If you’re reaching out about startups, hiring, product, or careers, start with these.

• Ship While You Sleep (2014; v2 2018) – on hiring, velocity, and building product like code-as-cards.
• Amol’s Guide for Beginners – a curated best-of selection about getting into startups and learning to build.
• Amol’s Team IQ Memo – a short book about leading, measuring, and improving teams.
• Doctoral Dissertation – for the historically curious.

Columbia Courses

I teach Columbia College undergraduate courses at Columbia University:

• Venturing to Change the World – see the syllabus: PDF
• Making History Through Venturing – see the syllabus: PDF

If you’re asking about startups, venture, or careers, those syllabi are a strong signal of how I think about building and history.

Selected Posts

• Patterns in Student Startups
• What To Do Between Startups
• Lessons from Growth Hacking My Blog
• Tech Is Taking Over NYC

Follow / Subscribe

• Newsletter (and RSS)
• Podcasts
• Twitter/X
• LinkedIn
• WhatsApp Group
• Drwn (older archive)

If You Email Me

I genuinely like helping people. I also have a finite inbox.

The fastest path to a useful response:

1. Read something above and reference it.
2. Ask one specific question.
3. Show your thinking and what you’ve already tried.

“Pick my brain” usually means “please do my thinking for me.”
A better move is: show me your thinking and ask for one correction.

This page exists so that high-leverage conversations start from shared context.

Tune in

1. John Borthwick (Betaworks) Highlight: Media, art, and agencies at the heart of early NY tech, plus predictions for today’s ecosystem. URL: https://amol.sarva.co/john-borthwick-betaworks/
2. Owen Davis (Contour, NYC Seed) Highlight: How advertising powered early NYC tech and the post-dot-com transition to consumer and enterprise superstars. URL: https://amol.sarva.co/owen-davis-contour-and-nyc-seed/
3. Albert Wenger (Union Square Ventures) Highlight: USV’s evolution from consumer apps to deep tech, lessons from Delicious, Tumblr, Foursquare, and frameworks for crypto + AI. URL: https://amol.sarva.co/albert-wenger-union-square-ventures/
4. Dennis Crowley (Foursquare, Dodgeball) Highlight: From punk zines and DIY publishing to inventing location-based social networking and Foursquare. URL: https://amol.sarva.co/dennis-crowley-foursquare-and-dodgeball/
5. Ben Lerer (Thrillist, Lerer Hippeau) Highlight: Building Thrillist in the newsletter/blogger era and evolving into a powerhouse seed-stage VC firm. URL: https://amol.sarva.co/ben-lerer-thrillist-and-lerer-hippeau/
6. Alan Patricof (APAX, Greycroft, Primetime Partners) Highlight: Oral history of venture capital, iconic deals from Apple to AOL, and repeatedly reinventing firms. URL: https://amol.sarva.co/alan-patricof-apax-greycroft-and-primetime/
7. Fred Wilson (Union Square Ventures) Highlight: Flatiron Partners, USV’s cultural roots, and investments like Etsy, Kickstarter, and SoundCloud. URL: https://amol.sarva.co/fred-wilson-union-square-ventures/
8. Howard Morgan (Renaissance, Idealab, First Round, BCap) Highlight: From mainframes and early computing to founding Renaissance and Idealab, shaping quant finance and New York’s tech ecosystem. URL: https://amol.sarva.co/howard-morgan-first-round-and-bcap/
9. Kevin Ryan (DoubleClick, Gilt, MongoDB, AlleyCorp) Highlight: Building DoubleClick into an ad-tech giant, then founding Gilt, MongoDB, and AlleyCorp to seed NYC’s startup scene. URL: https://amol.sarva.co/kevin-ryan-doubleclick-gilt-mongodb-alleycorp/
10. PJ Parson (Northzone, Spotify) Highlight: Scandinavian venture capital pioneer, scaling companies like Spotify, and bridging European and global tech. URL: https://amol.sarva.co/pj-parson-northzone-izettle-storytel/

Download it here, Amol’s Team IQ Memo.

From 1775 onward, the recurring plot twist is this: the power that best manages coalitions and finance usually beats the power that wins battles. Britain turned that into an art form; Napoleon turned the opposite error into performance art.

1775–1815: Britain’s two superpowers were credit and patience

1775–1783 (American Revolution): Britain loses not because it can’t fight, but because it’s fighting an expeditionary war across an ocean against a rebellion that can shop for patrons. France enters, then Spain and the Dutch—suddenly Britain isn’t suppressing a colony; it’s playing global whack-a-mole. The “lesson” rhymes with later eras: one local conflict becomes expensive when it becomes a coalition magnet.

1793–1815 (Revolutionary/Napoleonic Wars): Britain’s signature move is not a brilliant cavalry charge; it’s writing checks, building fleets, and keeping the coalition in the game even after allies get smashed. Austria loses? Rebuild it. Prussia collapses? Subsidize it. Russia sulks? Sweeten the deal. Britain weaponizes:

Maritime control (blockade, trade interdiction) Finance (credit credibility, subsidies) Industrial scale (early industrial advantage) Time (outlasting continental bursts of genius)

Napoleon’s misadventures are basically a tutorial in how not to handle third parties:

The Continental System (economic blockade of Britain) is the prototype of “sanctions as grand strategy”… and it backfires because enforcement requires coercing neutrals and semi-allies. When you try to control everyone’s trade, you end up at war with everyone’s pride.

Spain/Portugal (Peninsular War): he turns a manageable strategic problem into a legitimacy and guerrilla sinkhole by imposing a dynastic solution. It’s the recurring error: treating nations as tiles on a board instead of organisms with immune systems.

Russia 1812: the archetypal overreach—logistics + distance + scorched earth + winter + nationalism. But the deeper mistake isn’t weather; it’s believing one decisive campaign can end a multi-actor struggle. Coalitions don’t die when you beat an army; they die when you remove the reason for coalition.

Britain’s masterstroke: never lets Napoleon create a stable settlement. It keeps the coalition’s “why” alive, and funds the “how.”

1815–1914: The “Concert” era and the industrial ratchet

After 1815, Europe’s elites try to dampen the coalition-magnet problem with norms and congresses. It works… until industrialization accelerates everything: rail mobilization, mass armies, steel, chemistry. War becomes less “campaign” and more “national metabolism.”

This matters because it sets up 1914’s key pattern: industrial systems + rigid alliances + misread signals = cascade.

1914–1945: The Napoleon-to-Hitler rhyme

Germany and Japan repeat the Napoleonic sin in modern form: they pick fights that unify richer industrial systems against them.

Germany’s fatal pattern resembles Napoleon’s:

Enforce economic warfare (submarines) → anger neutrals Seek decisive outcomes → trigger wider coalition entry Overreach eastward → logistics + attrition + depth (Russia again, the eternal anvil)

Japan’s version is even cleaner: attack the U.S. to avoid strangulation, thereby guaranteeing the largest industrial mobilization on Earth joins the war.

The deep continuity from 1775 to 1945 is: wars are often lost the moment a challenger convinces outsiders that neutrality is unsafe or dishonorable.

Now: U.S.–China–Europe–Russia as a 21st-century remix

Here’s the “Napoleon/UK lens” applied to today:

The U.S. is playing Britain’s old role (coalition hub + finance/technology gatekeeper + maritime reach), but with a twist: the U.S. is also tempted to be a sheriff, not just a coalition manager.

China is tempted by the Napoleonic error: not “invade Russia in winter,” but coerce the trading system in ways that turn hedgers into enemies. The modern Continental System is any strategy that makes third parties feel their economic sovereignty is being rewritten by force.

Europe is the swing organism, not a pawn. Europe’s instinct is British-in-reverse: preserve commerce, reduce dependence, avoid binary alignment. If either superpower treats Europe like a compliance province, Europe’s immune response is “strategic autonomy” (sometimes theater, sometimes real).

Russia is a spoiler with depth, not a peer with surplus. In Napoleonic terms, Russia plays “strategic retreat + endurance + disruption.” It doesn’t need to “win” in a clean way; it needs opponents to bleed time, attention, and unity.

The two modern “misadventure” pathways (built from 1775–1812 logic)

Coalition-forcing economic warfare If the U.S. uses secondary sanctions/export controls in ways that look like Continental-System enforcement—i.e., “we control your trade too”—it risks pushing allies into building alternative rails. That’s not Europe “joining China,” but it is Europe refusing Britain-style subsidy alignment. A decisive-campaign fantasy If China believes a short, sharp action (Taiwan or coercive blockade-style pressure) can settle the question without producing a durable coalition, it risks replaying 1812 in modern form: not winter, but compounding logistics shocks, capital flight, and a long tech/finance siege that turns a tactical win into strategic anemia.

What “victory” looks like in the Britain-vs-Napoleon frame

Britain’s definition of victory: keep the coalition together, keep credit intact, keep the sea lanes and money lanes working, and let time do violence on your behalf. Napoleon’s definition of victory: force a settlement through decisive outcomes and system-wide control.

Translated to now:

U.S. durable win: allies cooperate because it’s advantageous and dignified, not because they’re threatened into compliance; tech and finance tools are used surgically so the coalition doesn’t fracture. China durable win: achieve security objectives without producing moral clarity + unified economic siege; keep third parties trading enough that containment never fully gels.

The uncomfortable punchline is that the greatest powers most often lose by trying to make the world legible and controllable. The world is not legible; it’s a swamp full of proud mammals with central banks.

If you want to go one level deeper, the cleanest “Napoleonic moment” to map onto the present is the Continental System—economic compulsion that turns bystanders into antagonists—and the cleanest “British moment” is subsidized coalition management—paying, sharing tech, and giving partners agency so they stay in the fight (or in the alignment) for years.

Now power is mostly about moving symbols and instructions: capital, code, designs, patents, standards, narratives. Tonnage still matters, but it’s no longer the headline act.

So what happened to rail–river–steel in this region?

They didn’t stop mattering. They got demoted from king to queen.

The Rhine is still Europe’s busiest freight corridor. Barges move chemicals, machinery, fuel, grain—quietly, cheaply, predictably. That stability is a kind of power: it lowers the cost of everything downstream. Basel to Rotterdam is an industrial bloodstream. Not glamorous, but you faint without it.

Rail and road here are less about invasion routes and more about just-in-time continental metabolism. Car parts cross borders five times before becoming a car. Pharma precursors bounce between Switzerland, Germany, and France like pinballs. That interdependence is deliberate: it makes war not just immoral but catastrophically expensive in spreadsheet terms.

Steel? Europe doesn’t dominate bulk steel anymore, but it dominates high-spec materials: specialty alloys, precision machine tools, industrial robotics. The value shifted from volume to tolerances. Power now hides in micron-level reliability, not blast furnaces.

Now to your sharper point: modern geopolitical power rests more on population, education, cohesion, innovation, and resources. That’s the real upgrade.

Population: Europe is aging, which is a strategic vulnerability. This region compensates with migration and dense urban networks, but demography is destiny with a long fuse.

Education: still elite. Universities, apprenticeships, research labs, and the weirdly effective German vocational pipeline create deep technical competence, not just abstract brilliance.

Social unity: mixed. The EU project turned former battlefields into shared administrative zones, which is historically miraculous. But cultural cohesion is thinner than economic integration, and politics can still get spicy when growth slows.

Technological innovation: Europe is better at systems engineering and regulation-as-power than platform capitalism. It writes rules that become global defaults: chemical safety, privacy, medical standards. That’s soft power with teeth. Not flashy, but it shapes markets.

Natural resources: weak on energy, stronger on water, arable land, and industrial ecosystems. Hence the obsession with supply chains and green transition—energy independence is the new Maginot Line, only this time the enemy is thermodynamics.

Here’s the twist that usually gets missed:

old infrastructure amplifies new forms of power.

High human capital plus terrible logistics is a science fair.

High human capital plus flawless logistics becomes economic gravity.

Innovation still needs factories, data centers, clean water, and reliable transport. Chips get designed in clean rooms but still ride trucks. Biotech runs on centrifuges and cold chains. Even cloud computing has a physical address and a cooling bill.

So this region’s historical advantages—dense transport, predictable borders, layered institutions—became the chassis on which the modern European project bolted itself.

War used to be about who could move armies fastest through this corridor.

Power now is about who can coordinate millions of peaceful transactions with the least friction.

Same map. Different game.

A thought experiment worth chewing on: if AI, automation, and energy storage keep advancing, does geography keep losing relevance—or does it swing back, with water access, grid stability, and climate resilience becoming the new strategic choke points? The Rhine might yet have another act, just wearing a lab coat instead of a uniform.

My favorite naive comment from people about Morrissey who know only two things is the second one (the first involves ‘mope’) — Morrissey’s politics are offensive to someone somewhere, mainly about the “England for the English” line and remixes thereof in song and comment (“Bengali In Platforms”…”life is hard enough when you belong here.”) Well, interpret the artist and the songs and the public performances…

Here are some other views you may know he holds:

• “The Queen is Dead” – the 80s album and song, pitched against the monarchy (AKA the Imperialist Colonialist Regime that terrorized the world for 300 years till they collapsed). He moved out of England entirely for LA about 20 years ago, renouncing the place
• “Meat is Murder” – animals as members of our moral community, also anti-bullfighting (“The Bullfighter Dies”); he makes venues change their concession menus or he won’t play
• Against the Iraq war and Middle East meddling by the West (“World Peace is None of Your Business”)
• Defending the latino immigrant populations in the US
• Critique of capitalism and consumerism (“Glamorous Glue”), and the record industry (“Paint a Vulgar Picture”)
• Organized religion (“Vicar in a Tutu”, “I Have Forgiven Jesus”)
• General thuggery – he’ll walk offstage if the venue security roughs up his flower-wielding fans

Oddly his 2014 album still isn’t on streaming, since he lame-o label of the time bungled it. It’s very good and here is what “World Peace is None of Your Business” is about. He has called it his BEST album.

A single argumentative essay disguised as a glam-rock record: a cranky Enlightenment pamphlet, a tabloid, and a Wildean epigram book all stitched together with brass riffs and bile. Morrissey isn’t really writing “protest songs” so much as moral cartoons — exaggerated figures used to poke at hypocrisy, cruelty, sentimental politics, and the theater of power.

1. World Peace Is None of Your Business

Core idea: Moral exhaustion with global virtue signaling.

This isn’t isolationism so much as ethical triage. Morrissey is skewering the modern posture of public empathy as performance: everyone has opinions on distant suffering, but nobody cleans up their own moral mess. It echoes Swiftian satire — the voice isn’t necessarily endorsing indifference, it’s exposing how compassion becomes a brand.

Politically, it brushes against post-Iraq/Afghanistan Western fatigue, the era of online outrage cycles, and NGO moral theater. Philosophically it smells faintly of Schopenhauer: suffering is endless; moral grandstanding doesn’t fix metaphysics.

Artistically, the brass arrangement gives it a carnival-barker tone — serious accusation delivered with mischievous pomp, like a political cartoon that grins while biting.

2. Neal Cassady Drops Dead

Core idea: The myth of American freedom crashing into mortality.

Neal Cassady was the kinetic muse of Kerouac and the Beat generation — speed, sex, movement, rebellion. Morrissey treats him less as hero than as a cautionary fossil: romanticized freedom ending in banal decay.

Literary reference is explicit: the Beats as American secular saints whose transcendence curdled into addiction and burnout. It quietly critiques the modern startup-ish cult of “move fast, break everything” before that phrase existed.

Ethically, it questions whether romanticizing self-destruction is just aestheticized nihilism wearing leather jackets and poetry.

3. Istanbul

Core idea: Erotic geopolitics and cultural projection.

“Istanbul” operates as orientalist fantasy turned inside-out. Rather than celebrating exoticism, Morrissey dramatizes how Western desire projects fantasies onto other cultures — danger, sensuality, romance, escape.

Politically it hints at border anxiety, migration, and Europe’s long obsession with the “edge of civilization.” Istanbul becomes a symbolic hinge between identities rather than a real city.

Artistically, it’s camp travel writing: part tourist postcard, part fever dream, part self-mockery. The exaggeration is the point.

4. I’m Not a Man

Core idea: Masculinity as costume rather than essence.

This is a gender song without being a culture-war tract. Morrissey interrogates masculinity as performance pressure — dominance scripts, emotional repression, tribal signaling — rather than biology.

Ethically it aligns more with existentialism than identity politics: the self isn’t a fixed category, but a negotiation with social expectations. There’s a faint Wildean refusal of compulsory normality.

Politically it brushes the edges of contemporary gender debates but refuses to be programmatic — more psychological than ideological.

5. Earth Is the Loneliest Planet

Core idea: Cosmic loneliness and anthropocentrism.

This track leans philosophical rather than political: humans believing themselves special while simultaneously isolated and self-destructive. It echoes late-romantic melancholy filtered through modern cosmic awareness — we’re a tiny species having oversized moral drama.

There’s a subtle ecological undertone: humanity as a lonely tenant mismanaging its only apartment.

Artistically it recalls Bowie’s space-age alienation phase more than classic Moz miserabilism.

6. Staircase at the University

Core idea: Intellectual cruelty and institutional vanity.

Universities here function as symbolic theaters of ego, hierarchy, and quiet humiliation. It’s not anti-intellectual so much as anti-pretension: knowledge systems producing status games instead of wisdom.

Literary lineage: the campus novel tradition (Kingsley Amis, David Lodge) plus Morrissey’s usual sympathy for outsiders humiliated by polite society.

Ethically it critiques how institutions sanitize cruelty under civility.

7. The Bullfighter Dies

Core idea: Ritualized cruelty disguised as tradition.

Bullfighting stands in for any cultural practice defended as “heritage” while causing suffering. Morrissey uses theatrical irony rather than literal activism — cruelty framed as spectacle until death punctures the fantasy.

Politically this dovetails with animal rights ethics, but the deeper question is: when does tradition become moral inertia?

Artistically it’s melodrama as moral fable — almost operatic in its moral clarity.

8. Kiss Me a Lot

Core idea: Erotic dependency and emotional bargaining.

This one feels personal rather than political: intimacy as power exchange, affection as currency, desire tangled with insecurity. It’s classic Morrissey terrain — romantic masochism filtered through adult fatigue rather than youthful melodrama.

Literary echoes of doomed romance archetypes rather than social critique.

9. Smiler With Knife

Core idea: Politeness masking violence.

A social psychology song. The smiling face conceals predation — corporate, romantic, political, social. It aligns neatly with modern distrust of “nice” branding, PR language, soft authoritarianism.

Ethically it argues that cruelty increasingly travels disguised as charm rather than brute force.

This is one of the album’s sharper metaphors — almost Orwellian in its distrust of pleasant surfaces.

10. Kick the Bride Down the Aisle

Core idea: Marriage as power ritual, not romance.

This is savage satire of heteronormative ceremony, social coercion, and romantic mythology. Wedding culture becomes spectacle masking domination, conformity, and transactional union.

Literary lineage: anti-marriage satire going back to Wilde and Shaw — the wedding as social theater rather than love story.

Politically it flirts with anti-institutional critique without preaching.

11. Mountjoy

Core idea: Incarceration, control, and invisible suffering.

Mountjoy Prison in Dublin anchors the song in real carceral geography. The track gestures toward institutional punishment, invisibility of prisoners, and moral distance between society and its locked-away populations.

Ethically it questions punitive justice rather than reform. It’s understated social critique rather than sloganeering.

12. Oboe Concerto

Core idea: Art as refuge against vulgarity.

Instrumental title evokes classical refinement, emotional interiority, and the idea that beauty still matters in a coarse world. Even when lyrics enter, the framing is about sensitivity versus brutality.

Artistically it’s Morrissey defending aesthetic seriousness — beauty as moral resistance rather than escapism.

TESLA AT ~300X P/E — HISTORICAL CONTEXT WITH AMAZON, APPLE, MSFT, NVIDIA

(All text preserved, no rich formatting, headings in CAPS)

AMAZON (AMZN)

Peak valuation moment:

Peak P/E (TTM): approximately 563x Dec 2015 quarter: P/E 563.23 with price $33.79, EPS $0.06 (cite: turn14search0) Approx market cap then: about $320B (2015 year-end) (cite: turn6search16) Net income in 2015: $0.59B (cite: turn8search3, turn8search10)

So the market was paying roughly 540–560x earnings on a sub-$1B profit base.

Subsequent earnings growth:

Net income 2015: $0.59B Net income 2020: $21.3B (cite: turn8search10) 5-year net income CAGR (2015→2020): about 105% per year

This is extraordinary compounding: earnings increased roughly 36x in five years.

Takeaway: At its craziest P/E, Amazon ultimately grew into a significant portion of the valuation — helped by an extremely tiny starting earnings base. The absolute dollar increase was huge: from $0.6B to more than $21B.

APPLE (AAPL)

Apple never traded anywhere near Tesla’s ~300x P/E. Its high is modest by comparison.

Peak valuation moment (recent decade):

Peak P/E (TTM, last 10 years): about 40x Dec 2024 quarter: P/E 40.44, price $255.59, EPS $6.32 (cite: turn3search7) Net income FY 2024: $93.736B (cite: turn12search0, turn12search3, turn12search6) Approx market cap at that time: around $3T (rising toward $4T by 2025) (cite: turn13search23, turn13search4)

So this is about 40x earnings on roughly $94B of profit.

Subsequent earnings growth (so far):

Net income 2024: $93.736B Net income 2025: $112.01B (cite: turn12search0, turn12search3, turn12search6) 1-year growth from 2024→2025: about 19.5% (Macrotrends cites this)

Takeaway: Apple’s “high” P/E is not remotely Tesla territory, and it rests on enormous earnings.

MICROSOFT (MSFT)

Peak valuation moment (recent decade):

Peak P/E (TTM, last 10 years): about 48x Dec 2017 quarter: P/E 48.06, price $85.54, EPS $1.78 (cite: turn3search12) Net income FY 2017: $21.2B (cite: turn9search10) Market cap at end 2017: around $600B (statmuse range about $605–610B) (cite: turn6search10, turn6search2)

So roughly 48x earnings on more than $20B of profit.

Subsequent earnings growth:

Net income 2017: $21.2B Net income 2022: $72.7B (cite: turn9search10) 5-year net income CAGR (2017→2022): about 28% per year

Takeaway: Paying roughly 50x for Microsoft at the moment cloud inflected worked very well, with substantial earnings expansion on a large base.

NVIDIA (NVDA)

Closest analogue to an “insane multiple that later proves sane,” thanks to the AI boom.

Peak valuation moment (recent decade):

Peak P/E (TTM, last decade): about 139x Apr 2023 quarter: P/E 138.75, price $27.75 (post-split), EPS $0.20 (cite: turn10search0, turn10search9) Net income FY 2023: $4.368B (cite: turn11search10, turn11search4) Market cap around mid-2023: just under $1T (Nvidia crossed the trillion mark that year) (cite: turn13search21)

So roughly 140x earnings on about $4–5B of profit ahead of the AI explosion.

Subsequent earnings growth:

Net income 2023: $4.37B Net income 2024: $29.76B Net income 2025: $72.88B (cite: turn11search10, turn11search4)

2-year net income CAGR (2023→2025): about 308% per year

Takeaway: High P/E was justified only because earnings detonated upward even faster.

TESLA TODAY IN CONTEXT

Tesla’s current P/E (~300x) is being placed on tens of billions of earnings. Historically:

Amazon hit 560x but on less than $1B of earnings. Nvidia hit ~140x but on ~$4B of earnings. Microsoft and Apple have peaked around 40–50x on $20–100B+ of earnings.

The pattern:

Extreme P/Es only “work” when the earnings base is tiny and explodes afterward (Amazon, Nvidia). Mature megacaps with huge profits may justify 40–50x P/E when they post consistent double-digit earnings compounding (Apple, Microsoft). Tesla at 300x on a large earnings base requires Amazon- or Nvidia-level CAGRs on enormous profit dollars. That implies turning maybe $20B of earnings into something like $500B+ within a decade. It is mathematically possible but economically extraordinary.

Vs the US (what would you zero out?)

Federal spending is about 23–24% of U.S. GDP in recent years. Think of it as a big, lopsided pie with a fondness for retirees and missiles. The rough allocation looks like this:

Elderly:

Around 10% of GDP.

This includes Social Security (~5% GDP) and Medicare (~5% GDP). America’s fiscal metabolism is geriatric-centric; these two dominate everything else.

Poor / Income Support:

Roughly 2–3% of GDP.

This covers programs like SNAP, the Earned Income Tax Credit, housing vouchers, TANF, SSI for the disabled poor, and other means-tested transfers. It’s much smaller than political rhetoric suggests.

Sick (non-elderly health care):

About 2% of GDP.

Primarily Medicaid for non-elderly recipients, plus ACA subsidies. Medicaid is half elderly/disabled and half low-income adults and kids; this bucket is the non-elderly portion.

Security / Defense:

Roughly 3% of GDP.

This is the Pentagon’s world: personnel, procurement, R&D, operations. Add another ~0.3% for homeland security and veterans’ health if you want a more expansive definition.

“Everything Else”:

Around 5–6% of GDP.

This grab-bag covers the civil service, transportation, science, education, diplomacy, agriculture, energy, courts, and the machinery that keeps a continental republic humming—basically the part everyone forgets exists.

The last 20 years were a lot of talk about Vitamin D. In the end, it’s pointless to take it (though it doesn’t hurt, except your pocketbook.)

Scientific American roundup and deep dive article.

https://archive.ph/XuZ40

Taurine’s Interaction with GABAergic Signalling

Taurine is a β-amino sulfonic acid abundant in brain tissue and acts as a neuromodulator. It binds and activates certain subtypes of the GABA_A receptor (ionotropic inhibitory receptor) in the human/mammalian brain, although with lower potency than GABA itself — for example, EC₅₀ ≈ 116 µM for taurine vs ~3.7 µM for GABA in one study.  Taurine’s activation is subunit-dependent: it preferentially activates extrasynaptic GABA_A receptors containing subunits such as α4β2δ, which mediate tonic inhibition (constant, low-level inhibition) rather than fast synaptic phasic inhibition.  In the thalamus, taurine application in brain slices reduced neuronal firing and input resistance, through tonic currents via extrasynaptic GABA_A receptors — revealing its physiological modulatory role of dampening excitability. 

Chronic taurine administration in animal models influences GABAergic system regulation: it increases the expression of glutamic acid decarboxylase (GAD, enzyme making GABA), and may shift GABA_A receptor subunit composition (β2/β3) to favour chloride influx/hyperpolarisation.  Because GABA is the major inhibitory neurotransmitter in the CNS, taurine’s effect via GABA_A modulation supports neuroprotective, sedative/a-anxiolytic functions rather than classic stimulant effects.

For example, researchers found that taurine may make the brain feel more “calm” rather than “wired”, which might paradoxically contribute to energy-drink “crash” phenomena when co-ingested with stimulants. 

Implication: Supplementing taurine likely enhances inhibitory signalling in neural circuits by engaging GABA_A receptors (especially extrasynaptic subtypes), which can moderate excitability, may reduce jitter/over-activation, support neuroprotection, and modulate mental-state tension. But it does not act as a direct stimulant for focus/alertness in the classical sense — its primary role via GABA is dampening rather than activating.

Abbreviated Summary of Other Findings

Energy & Focus: Human trials show modest physical-endurance improvements (e.g., longer time-to-exhaustion) with taurine (1–6 g doses) mainly in exercise contexts; cognitive focus/attention benefits from taurine alone are weak or null.

Longevity & Aging: Observational human population data show higher taurine levels correlate with lower incidence of cardiovascular/metabolic disease and better health in older adults; animal data show taurine supplementation extends lifespan (~10–12 % in mice).

Safety: Taurine supplementation (commonly up to ~3 g/day) is widely well-tolerated in adults, with minimal side-effects. Caution in children, pregnancy, or high-dose long-term usage due to limited data.

Source URLs (GABA / Taurine Mechanism)

https://pmc.ncbi.nlm.nih.gov/articles/PMC5933890/ https://www.frontiersin.org/journals/pharmacology/articles/10.3389/fphar.2023.1271203/full https://pubmed.ncbi.nlm.nih.gov/30892104/ https://pmc.ncbi.nlm.nih.gov/articles/PMC6671153/ https://news.cornell.edu/stories/2008/02/scientists-close-taurines-activity-brain https://jbiomedsci.biomedcentral.com/articles/10.1186/1423-0127-17-S1-S14 https://journals.lww.com/nrronline/fulltext/2024/01000/emergence_of_taurine_as_a_therapeutic_agent_for.10.aspx

The Harvard/MIT crowd I read says: the Tylenol link is weak. Moms take Tylenol because they’re sick, and it’s the fever itself that’s more strongly tied to ASD risk. That’s why the real research is focused on immune pathways, not acetaminophen.

Folate and leucovorin? Same story—no meaningful data.

Investors and scientists who know the field are mostly ignoring this cycle. The political promises about “new datasets” never really landed, and the underlying science hasn’t budged. The good news is more attention may still push more funding.

For families, though, it’s a mess. Partial theories, half-baked media takes—it’s confusing and cruel when people are desperate for clarity.

The work continues where it always has: immunology, genetics, neurodevelopment. That’s the real path.

Links if you want more:

• Stat News

• Spectrum / Simons Foundation

• Endpoints

And much more—

Causes of Autism Spectrum Disorder and Promising Interventions

Introduction

Autism spectrum disorder (ASD) is a complex neurodevelopmental condition with diverse presentations. Its causes have been the subject of intense research and public speculation. Scientists now view ASD as arising from a combination of genetic predispositions and environmental influences, which together affect early brain development  . While the exact mechanisms remain only partially understood, a dominant consensus has emerged on several key contributors. In parallel, a range of therapeutic and interventional strategies – from behavioral therapies to cutting-edge biomedical approaches – are being explored to support individuals on the autism spectrum. Below, we review the current scientific understanding of ASD’s causes (biological, genetic, environmental, and immunological) and evaluate specific claims about risk factors like Tylenol, folate, leucovorin, and maternal fever. We also highlight promising interventions and research directions, distinguishing well-supported findings from speculation.

Genetic and Biological Contributors to ASD

Research consistently shows that genetic factors are the primary contributors to ASD risk. Twin and family studies indicate autism is highly heritable, with genetics accounting for an estimated 64–91% of variance in susceptibility . Over the past decade, scientists have identified hundreds of genes associated with ASD. The Simons Foundation’s SFARI database, for example, now lists around 942 candidate genes implicated in autism, including about 100 high-confidence risk genes strongly linked to the condition . These genes vary in effect: some are rare mutations with large impact (often arising spontaneously, or de novo, in the child), while others are common inherited variants that each confer small risk but can act additively . Together, these genetic factors influence brain development and neural connectivity, aligning with biological findings that autistic brains show differences in how neurons form connections and communicate.

Importantly, autism’s genetic architecture is heterogeneous – there is no single “autism gene.” Known ASD-related genes converge on a few biological pathways, notably those involved in synapse formation, neuronal signaling, and gene regulation in the developing brain . For example, certain mutations affect proteins that help neurons connect (potentially altering brain circuit wiring), while others affect when and how genes are turned on in brain cells. In some cases, a genetic change can be directly causal: well-known genetic syndromes like Fragile X syndrome, Rett syndrome, Tuberous Sclerosis, or PTEN-associated conditions often include autism as a feature. These illustrate how biological alterations (e.g. disrupted synaptic proteins or increased cell growth signaling) can lead to autism-like neurodevelopmental differences.

On the neurological level, studies of autistic individuals have found atypical patterns of brain growth and activity. Some infants later diagnosed with autism show unusually rapid brain growth in the first years of life, followed by differences in connectivity between brain regions. Functional imaging studies often find altered communication between brain networks involved in social communication and sensory processing. These biological observations, while not yet specific enough for diagnosis, reinforce that ASD arises from early differences in brain development – differences largely set in motion by genetic and molecular factors.

Bottom line: The scientific consensus is that genetics and neurobiology play a central role in autism. In the majority of cases, inherited or de novo genetic variations predispose a child’s brain to develop differently, creating the spectrum of traits we recognize as ASD. However, genetics alone is not the whole story; researchers also study how environmental and immune factors might interact with these predispositions.

Environmental and Prenatal Risk Factors

While genes create a vulnerability to ASD, environmental factors during prenatal and early life can also contribute (albeit to a lesser extent). In this context, “environmental” means any non-genetic influence – from parental age to exposures in the womb . It’s crucial to note that no single environmental agent causes autism by itself; rather, these factors may increase risk modestly or trigger ASD only in combination with genetic susceptibilities  .

Established associations include:

• Advanced parental age: Older parents, especially fathers, are consistently linked to higher ASD likelihood . This may relate to a greater chance of gene mutations in sperm/eggs as age increases.

• Birth complications and prenatal factors: Extremely premature birth, very low birth weight, and prenatal factors like lack of oxygen (perinatal asphyxia) have been associated with later ASD. These stressors can affect early brain development.

• Maternal conditions: Maternal obesity and diabetes during pregnancy have shown modest correlations with autism in some studies. Researchers speculate that metabolic or inflammatory effects in the womb could play a role, though these links are not fully consistent.

• Certain drug exposures: One well-documented example is the anti-seizure medication valproic acid. If taken during pregnancy, valproate substantially increases the risk of autism and related neurodevelopmental disorders in the child . It likely interferes with fetal neural development (which is why it’s now cautioned against in pregnancy). Another historical example: prenatal exposure to thalidomide (a drug once used for morning sickness in the 1960s) was found to cause birth defects and a higher incidence of autism in a small cohort – timing of exposure in early pregnancy appeared critical. These cases, while rare, illustrate that powerful teratogens can impact pathways relevant to autism.

• Prenatal infections and inflammation: Maternal infection during pregnancy – and the resulting immune response – has emerged as a significant environmental risk factor (discussed more under Immunological Factors below, including maternal fever). In utero exposure to infections like rubella (German measles) is a known cause of autism in a notable fraction of cases; prior to widespread rubella vaccination, outbreaks led to spikes in autism rates among exposed children. This and other data support the idea that maternal immune activation can alter fetal brain development in ways that sometimes lead to ASD.

Beyond these, dozens of other environmental suspects have been investigated. These include prenatal exposure to air pollution (e.g. traffic-related smog or pesticides), maternal prenatal stress, nutritional factors, method of childbirth (C-section vs vaginal), spacing between pregnancies, and prenatal use of certain medications like SSRIs (antidepressants). Results have often been inconsistent or conflicting  . For instance, some studies initially suggested that prenatal SSRI antidepressant use might increase autism risk, but later analyses pointed to maternal depression (the reason for SSRI use) as the confounding factor rather than the medication itself. Similarly, early reports that babies born shortly after an older sibling had higher autism rates (implying a short inter-pregnancy interval risk) were not uniformly replicated. Environmental epidemiology in autism is notoriously challenging because it’s hard to separate correlation from causation in observational studies . Families with one child on the spectrum may share genetic or lifestyle factors that also affect their other children, muddying the waters on environmental correlations.

Crucially, some high-profile environmental theories have been thoroughly investigated and refuted. The most well-known is the idea that childhood vaccinations (such as MMR) cause autism – a notion stemming from a discredited 1998 study. Extensive research involving millions of children has found no credible link between vaccines and ASD . Large-scale studies in multiple countries have shown that autism rates are the same in vaccinated and unvaccinated children . Likewise, concerns about the vaccine preservative thimerosal (a form of mercury) were not borne out by data – removing thimerosal from vaccines did not reduce autism rates . The scientific consensus is unequivocal that vaccines do not cause autism, and the initial vaccine-autism claim is now regarded as an example of how false leads can distract from real progress.

In summary, environmental factors can contribute to ASD, but usually in combination with genetic risk. Among the myriad factors studied, only a few (like parental age and certain prenatal exposures) have consistent evidence . Even these increase risk only slightly (on the order of a 1.5-fold to 3-fold increase in odds, not a deterministic cause). The difficulty in pinning down environmental causes lies in separating true causative effects from coincidental associations and confounders . Researchers emphasize that observing an association (e.g. more autism in children of mothers who experienced X) does not prove causation – rigorous methods (controlled studies, animal models, etc.) are needed to confirm if and how an exposure directly impacts neurodevelopment.

Immunological Factors and Maternal Influences

One intriguing piece of the autism puzzle involves the immune system – both the mother’s immune environment during pregnancy and the child’s own immune function. While this line of research is ongoing, there is evidence that immune-related events in early development can increase the likelihood of ASD in some cases.

Maternal immune activation: When a pregnant woman’s immune system is strongly activated – for instance, by a significant infection or sustained inflammation – it can affect the developing fetal brain. In both animal models and human studies, this phenomenon (often called maternal immune activation, or MIA) has been linked to autism-like outcomes. A notable example is maternal fever during pregnancy. Fever is a common immune response to infection, and studies have found that women who have frequent or prolonged fevers while pregnant are more likely to have children diagnosed with ASD  . Specifically, one large Norwegian cohort study reported that fever in the second trimester was associated with a 40% increase in autism risk (adjusted odds ratio ~1.4) compared to no fever, and mothers who had three or more fever episodes after the 12th week of pregnancy had over 3 times the risk of ASD in their child  . These findings support a dose-dependent effect of immune activation on neurodevelopment. It appears that inflammatory molecules (cytokines) from the mother can cross into the fetal environment and potentially disrupt brain developmental processes at critical times . It’s important to note that infection itself is not “destiny” – most babies born to mothers who had infections or fevers are not autistic. Rather, maternal infection is a risk factor that may contribute in a subset of cases (possibly interacting with genetic vulnerabilities in the child).

Interestingly, treating maternal fever might mitigate some risk. The same research observed that mothers who used antipyretic medications (like acetaminophen) for their fevers had lower ASD rates than those who did not, hinting that reducing inflammation could be protective . (This must be balanced against recent questions about acetaminophen safety itself – covered in the next section.) Maternal vaccination against infections (flu, COVID-19, etc.) during pregnancy, which prevents severe illness and fever, is another indirect way to reduce this risk factor.

Maternal autoantibodies: Another immune-related contributor under investigation is the presence of maternal antibodies that react to fetal brain proteins. Normally, antibodies do not attack one’s own tissues, but in some people an autoimmune phenomenon occurs. Research has found that about 10–20% of mothers of children with autism have immune antibodies in their blood that bind to proteins found in the developing brain . These are often referred to as maternal autoantibody-related (MAR) autism. The hypothesis is that these autoantibodies, if they cross the placenta during pregnancy, could interfere with the fetal brain’s development. In animal studies, scientists have injected such antibodies into pregnant mice or monkeys and observed autism-like changes in the offspring’s behavior, suggesting a causal role. This is still an emerging field, but it has big implications: if a subset of ASD is truly caused by maternal autoantibodies, it might be possible to identify those cases with a blood test and even prevent them (for example, by plasmapheresis to remove antibodies or other immunotherapies during pregnancy). Clinical trials are a long way off, but the maternal autoantibody link is one of the more convincing immunological findings so far in autism research.

Child’s immune system and neuroinflammation: Beyond the maternal influences, some studies have observed that autistic individuals themselves may have differences in immune function. Postmortem analyses of autistic brains have, in some cases, found signs of neuroinflammation – such as activated microglia (the brain’s immune cells) or elevated inflammatory cytokines in cerebrospinal fluid. Peripheral immune differences (like atypical levels of immune cells or antibodies in the blood) have also been reported in groups of children with ASD. It’s not yet clear whether these immune findings are a cause or an effect of neurodevelopmental differences. One possibility is that certain genetic mutations associated with ASD also affect immune regulation (for example, some autism-related genes are expressed in immune cells or microglia). Another possibility is that chronic inflammation could exacerbate autism symptoms or severity. The immune-autism connection is complex and not fully worked out, but it underscores the idea that ASD is not purely a “brain condition” in isolation – it may involve broader physiological systems, including the immune system. This has prompted exploratory trials of anti-inflammatory or immune-modulating treatments in autism (such as targeted anti-inflammatory drugs or IVIG in small subgroups), though none are yet proven effective.

In summary, immunological factors – particularly maternal infections and autoimmunity – are recognized as part of the landscape of ASD risk factors. The dominant view is that these factors alone do not cause autism, but they can interact with genetic and developmental factors to increase risk or shape the course of brain development . As research progresses, we may better distinguish an “immune-related autism” subtype, which in the future could even lead to preventative strategies for those cases.

Examining Controversial Claims: Tylenol, Folate, Leucovorin, and Maternal Fever

Over the years, various claims linking certain maternal behaviors or exposures to autism have gained public attention. It’s important to evaluate each with scientific rigor, as the quality of evidence varies widely. Below we discuss four such topics – acetaminophen use in pregnancy, folic acid (folate) supplementation, leucovorin therapy, and maternal fever – and assess what current research says about their connection (or lack thereof) to ASD:

• Acetaminophen (Tylenol) Use During Pregnancy: Acetaminophen is a common pain and fever reliever recommended for pregnant women (since alternatives like ibuprofen are unsafe in pregnancy). In recent years, some studies have reported a statistical association between prenatal acetaminophen use and higher rates of autism or ADHD in offspring . For example, large observational cohorts (including the Boston Birth Cohort and others) noted that children born to mothers who frequently used acetaminophen were diagnosed with ASD at slightly higher rates . However, association is not causation. These studies have significant limitations: mothers who needed a lot of pain relief or had persistent fever/illness (necessitating acetaminophen) may differ in other ways that affect their child’s development. Indeed, more rigorous analyses controlling for genetic and familial factors have cast doubt on this link. A 2023 study of nearly 2.5 million Swedish children, which included a sibling-comparison design, found that when comparing siblings (one exposed to acetaminophen in utero, another not), there was no increase in autism risk attributable to acetaminophen  . In that dataset, the apparent association seen in simpler analyses disappeared under tighter control, suggesting confounding factors (like underlying maternal health or genetics) were responsible .

The scientific community is somewhat split on this issue. In late 2025, the U.S. FDA took a cautious stance, announcing it would update acetaminophen labels to mention a possible association with neurodevelopmental outcomes  . The FDA cited a “considerable body of evidence” of correlation, but also acknowledged that causality is unproven and studies are conflicting . In contrast, the American College of Obstetricians and Gynecologists (ACOG) responded strongly, stating that “not a single reputable study” has found that acetaminophen use causes autism and warning that discouraging its use could do more harm than good . ACOG pointed out that two of the highest-quality studies, including the sibling-controlled analysis in JAMA, found no link between prenatal acetaminophen and ASD  . They emphasize that untreated pain or high fever in pregnancy can pose real risks (e.g. fever itself is a risk factor for pregnancy complications and possibly ASD), and acetaminophen is one of the few safe options to manage those issues .

Bottom line: Current evidence does not establish that Tylenol causes autism. Some observational studies raise concern, but their validity is limited by confounding biases. Higher-quality research (and expert organizations) tend to find no significant effect or at most a very small one . The prudent advice is for pregnant women to use acetaminophen only as needed (which is good practice for any drug in pregnancy) but not to endure harmful fevers or pain out of fear of autism. The debate here exemplifies how a possible environmental risk must be weighed against its benefits – untreated maternal fever, for instance, clearly can endanger a fetus, whereas the “acetaminophen-autism” link remains unproven and, if real, would likely be a minor contributor relative to other factors.

• Maternal Folate (Folic Acid) Intake: Folic acid is a B-vitamin (vitamin B₉) critical for fetal neural development. Medical guidelines strongly advise women to take folate supplements in early pregnancy to prevent neural tube defects (like spina bifida). The question has arisen: could folate be related to autism risk, either protective or harmful? Some studies have suggested adequate folate might reduce autism risk. Notably, a large Norwegian study in 2013 found that mothers who took folic acid supplements around conception had lower odds of having a child with autism . This makes biological sense, as folate is vital for DNA synthesis and methylation, processes important in brain development. However, subsequent research has been mixed: a similarly large Danish study found no significant link between prenatal vitamin use and autism rates . So the idea of folic acid as protective remains plausible but not conclusively proven.

On the flip side, concern about excessive folate emerged from a Johns Hopkins-led study. Researchers measured mothers’ blood folate levels at delivery and found that mothers with extremely high folate (and vitamin B₁₂) levels had an increased chance of their child later being diagnosed with ASD  . Specifically, mothers in the highest folate range (folate ≥ 60 nmol/L in plasma) were about 2.5 times more likely to have a child with ASD compared to mothers with more moderate folate levels . Very high B₁₂ levels showed a similar 2.5× risk increase . It’s important to stress that these mothers’ folate levels were very high – in some cases, far above what one would get from normal prenatal vitamins (possibly due to genetic differences in vitamin metabolism or taking overly large doses of supplements). The researchers described a U-shaped relationship: moderate folate was associated with the lowest autism risk, whereas both low and excessively high folate levels were associated with higher risk  . They cautioned that this finding is hypothesis-generating and does not negate the importance of sufficient folic acid . In other words, mothers should absolutely take the recommended folate (to prevent serious birth defects); the question mark is only around super-doses or unusually high blood levels, which might reflect some metabolic anomaly.

Bottom line: Folate is beneficial for pregnancy, and if anything, normal supplementation may help lower autism risk (especially in high-risk situations like mothers on anti-epileptic medications – folate has been shown to reduce developmental issues in those cases). There is no good evidence that standard folic acid causes autism. One study noted an association between extraordinarily high folate/B₁₂ levels and ASD , but this does not mean folate is harmful – it suggests an area for further research (for instance, investigating if certain mothers over-supplement or have genetic traits leading to high folate retention). The scientific consensus remains that women should take the recommended prenatal folate, as its proven benefits (preventing neural tube defects and possibly other developmental problems) far outweigh any unconfirmed autism speculation.

• Leucovorin (Folinic Acid) Therapy: Leucovorin is a form of folate (vitamin B₉) known as folinic acid, which is more bioactive than standard folic acid. It’s used medically to “rescue” cells during certain chemotherapy and to treat folate metabolism disorders. In the autism context, leucovorin has drawn interest not as a causative factor but as a potential treatment for a subset of children. This stems from findings about cerebral folate deficiency and folate receptor autoantibodies in some individuals with ASD. A portion of autistic children have been found to carry antibodies that block folate transport into the brain (the aforementioned folate receptor autoantibodies, or FRAA). This can lead to low folate levels in the central nervous system despite adequate folate in the diet. High-dose folinic acid can bypass this blockage and raise brain folate levels.

In 2016, researchers conducted a randomized, placebo-controlled trial of high-dose leucovorin in children with autism (most of whom had language delays). The results were encouraging: 12 weeks of folinic acid (dose ~2 mg/kg per day) led to significant improvements in the children’s verbal communication skills compared to placebo . On a standardized language assessment, the treatment group improved about 5.7 points more than the placebo group on average (a medium-to-large effect size) . Notably, children who tested positive for FRAA (the folate-blocking antibodies) showed the greatest gains – about a 7-point improvement in communication, versus little change in placebo (a large effect size)  . These kids also showed improvements in other areas like daily living skills and reduced autism symptom severity on caregiver checklists . There were no significant side effect differences between leucovorin and placebo .

This study, though relatively small (48 children), provides proof of concept that targeting folate pathways can help some autistic children, particularly those with an identifiable metabolic/immune abnormality (FRAA). As a result, some clinics now test for folate receptor antibodies in kids with ASD and consider a trial of leucovorin if they’re positive. Larger trials are underway to confirm and extend these findings. It’s important to understand that leucovorin is not a general “autism cure” – it appears to benefit a specific subgroup with folate pathway issues. For children without those issues, extra folinic acid might have little effect. But this line of research exemplifies a trend toward precision medicine in ASD, where biological subtypes (like folate autoimmunity) could be treated with targeted interventions.

Bottom line: Leucovorin (folinic acid) has shown promise as a therapeutic intervention for a subset of children with ASD, especially those with folate receptor antibodies and language delays . This claim is supported by a controlled trial, making it one of the more valid “claims” in the autism treatment arena (in contrast to many unproven alternative therapies). It does not suggest that folinic acid deficiencies cause autism broadly, but rather that some children’s autism symptoms may stem in part from treatable folate metabolism problems. Parents should always consult a medical professional before pursuing such off-label treatments, as responses can vary and more research is ongoing.

• Maternal Fever: We touched on maternal fever under immunological factors, but it’s worth reiterating in this context because it directly ties into an actionable question for pregnant women: Does having a fever while pregnant increase autism risk, and what should be done about it? Epidemiological studies say yes, fever is associated with higher ASD risk, but with important nuances. As noted, fever, especially in the second trimester, appears to have about a 1.4-fold effect on risk . Multiple fevers amplify that risk (three or more fevers after the first trimester led to roughly triple the odds of ASD) . These data come from large prospective cohorts, lending them credibility. The mechanism is presumed to be maternal immune activation – essentially, the mother’s inflammatory response to infection, indicated by fever, may influence the fetal brain’s trajectory.

However, it would be wrong to conclude “fever causes autism” in a straightforward way. Many women have fevers (for example, due to common infections like the flu) and most of their children do not develop ASD. Fever is likely one risk factor among many, and it might need to coincide with other vulnerabilities (genetic or timing-wise) to have an effect. From a practical standpoint, these findings underscore the importance of managing fevers during pregnancy. Pregnant women are advised to promptly treat high fevers (with acetaminophen, which is safest for the fetus, and seek medical guidance for the underlying infection). There is even evidence that doing so can lessen any potential risk that the fever might pose to the baby’s neurodevelopment . Preventing infections in the first place is also key – for example, getting the flu vaccine and other recommended vaccinations during pregnancy reduces the chance of serious illness and fever, thereby indirectly protecting the fetus.

The maternal fever connection also fuels an area of autism research looking at anti-inflammatory interventions. If excessive inflammation during pregnancy is harmful to the fetal brain, could anti-inflammatory or immune-modulating treatments protect at-risk pregnancies? Thus far, this remains theoretical; the priority in clinical practice is simply to ensure infections are managed and fevers are treated.

Bottom line: Maternal fever is a real, if moderate, risk factor for ASD. This link is supported by epidemiological data  and aligns with a broader understanding that the intrauterine environment – particularly maternal immune activity – can influence brain development. Pregnant women should not panic about fevers per se, but they should take them seriously: consult healthcare providers, reduce the fever, and address the cause. By doing so, they not only improve their own comfort but also potentially reduce any associated risk to the developing baby.

Promising Research Directions and Interventions for ASD

Despite there currently being no “cure” for autism (and many autistic individuals and advocates caution against framing it solely as something to be cured), there are numerous interventions that can significantly improve quality of life and skills for people on the spectrum. Moreover, new therapeutic avenues are continually being explored. Here we summarize the landscape of behavioral, pharmacological, and biotech-driven interventions, highlighting both established approaches and emerging research.

Behavioral and Developmental Interventions

Behavioral and educational therapies have the strongest evidence for benefiting individuals with ASD, especially when started early. Decades of research have shown that early intensive behavioral intervention (EIBI) can improve language, cognition, and adaptive behavior in many young children with autism. One well-known approach is based on the principles of Applied Behavior Analysis (ABA) – a therapy that breaks skills into small steps, uses positive reinforcement, and is tailored to each child’s needs. Clinical studies and meta-analyses have found that some children in ABA programs make substantial gains in IQ, language, and daily living skills, though responses vary. The consensus recommendation (from the American Academy of Pediatrics, NIH, and others) is that children with ASD should receive structured, appropriate interventions as early as possible, ideally in the preschool years, to capitalize on the developing brain’s plasticity . In fact, routine autism screening at 18–24 months is done largely to identify kids who can benefit from early intervention services.

Specific early intervention models, such as the Early Start Denver Model (ESDM) – which combines ABA techniques with developmental “naturalistic” approaches for toddlers – have shown positive results in randomized trials. Other therapies focus on particular domains: speech and language therapy to foster communication (for nonverbal or minimally verbal children, this might include alternative communication methods like picture exchange or speech-generating devices), occupational therapy to address sensory and motor skills, and social skills groups to help older children and teens learn social rules in a structured way. These interventions are often combined into a comprehensive program. It’s common for young children with autism to have a team of providers – for example, a behavioral therapist working on reducing challenging behaviors and building new skills , a speech therapist focusing on communication, and an occupational therapist addressing sensory needs.

A key concept is that one size does not fit all in autism therapy. Individuals with ASD have a wide range of strengths and challenges, so intervention plans must be highly individualized  . Some children might need help primarily with communication, others with reducing self-injury or aggression, others with learning to cope with sensory sensitivities. Family involvement is also crucial: coaching parents to use strategies at home (for example, techniques to encourage communication or manage behavior) can greatly amplify the gains from therapy. Many interventions now are naturalistic, happening in the context of play or daily routines, rather than discrete drills at a table – this helps with generalization of skills.

For school-aged children, educational interventions become key. Under laws like IDEA in the US, children with ASD are entitled to support in school, such as individualized education programs (IEPs), special education services, or accommodations in mainstream classrooms. Behavioral strategies continue to be used in schools to help with classroom skills and peer interactions. For higher-functioning students, cognitive behavioral therapy (CBT) adapted for autism can be effective to manage anxiety or obsessive behaviors. As individuals enter adolescence and adulthood, interventions focus more on life skills, vocational training, and social coaching to facilitate independent living and employment.

The evidence base supports that intensive early intervention can significantly improve outcomes – some children who receive early therapy gain enough skills to later participate in mainstream classes, and a minority no longer meet the full criteria for ASD after years of intervention (though they may still have subtle differences). Even for those who remain substantially affected, building communication and self-care skills can vastly improve their autonomy and reduce frustration (for instance, being able to communicate one’s needs can decrease behavioral outbursts that stem from not being understood). The mantra in the field is “early and often”: engage children in as much beneficial therapy and education as reasonably possible, as early as possible  . This does not mean endless hours of rote therapy – it means giving them enriched learning environments and supports during the critical formative years.

In sum, behavioral interventions are the cornerstone of autism treatment. They are supported by decades of research and are currently the most effective tools we have to help individuals with ASD reach their full potential. Ongoing research in this domain is looking at how to tailor interventions to each child’s profile, how to involve technology (for example, using apps or robots to engage kids), and how to extend these services to more families (e.g. via telehealth coaching for parents in underserved areas).

Pharmacological Treatments and Trials

There is no medication that can “turn off” autism or address the core social communication challenges in a universal way. However, medications can be very helpful for managing certain symptoms or co-occurring conditions that many autistic people experience. The approach is often to treat specific issues (e.g. severe tantrums, inattention, anxiety) to improve day-to-day functioning and learning capacity.

Currently approved medications: The U.S. FDA has approved only two drugs specifically for autism: risperidone and aripiprazole. These are atypical antipsychotic medications, and they are approved not for autism’s core symptoms, but for treating irritability and aggressive behaviors often seen in youth with autism. Studies show that these medications can reduce severe tantrums, self-injurious behavior, and aggression in many children with ASD, which can greatly enhance quality of life for the child and caregivers. However, they carry side effects (weight gain, sedation, etc.), so they are usually reserved for more severe behaviors. Beyond these, clinicians commonly use other medications off-label to target specific symptoms: for example, stimulant medications or other ADHD drugs for children with significant hyperactivity and attention problems , SSRIs or anti-anxiety medications for those with anxiety, OCD-like symptoms, or depression , and sleep aids (like melatonin) for insomnia, which is very common in ASD. Each of these medications addresses a particular co-occurring issue rather than autism itself, but alleviating such issues can indirectly help a child engage better with therapies and learning.

Research on core symptoms: A major goal in autism research is to find medications or biological treatments that can improve the core social and communication difficulties or the repetitive behaviors and restricted interests. Several avenues have been tried or are under active investigation:

• Oxytocin and vasopressin pathways: Oxytocin is a hormone involved in social bonding, and early small studies hinted that oxytocin (often given as a nasal spray) could increase social attention or empathy in people with autism. This led to multiple trials; unfortunately, large placebo-controlled studies to date have had mixed results, with some failing to show a clear benefit. Research continues, including efforts to identify if certain subgroups (perhaps those with specific oxytocin receptor gene variants) respond better. A related approach involved blocking the vasopressin 1a receptor (vasopressin is another hormone linked to social behavior) – a drug called balovaptan showed promise in adults with ASD in early trials but ultimately did not show efficacy in a Phase 3 trial. Despite setbacks, the social hormone theory of treating autism is still being explored, and newer compounds or combinations might yet prove useful.

• Excitatory/Inhibitory (E-I) balance modulators: A prominent hypothesis is that autism involves an imbalance between excitatory and inhibitory signaling in the brain (too much “noise” or not enough inhibitory calming of neural circuits). Bumetanide, a diuretic, gained attention because it lowers intracellular chloride and can make inhibitory signals (via GABA) more effective in the developing brain. Several small trials suggested bumetanide might modestly improve social behaviors and reduce autism severity in some children, presumably by shifting the E-I balance. A meta-analysis found some positive effects, but a large European Phase 3 trial recently reported no significant benefit over placebo . It’s possible that bumetanide helps a subset or that longer treatment is needed, but its future is unclear given mixed evidence. Other drugs targeting the glutamate/GABA systems (the main excitatory/inhibitory neurotransmitters) have been tested – for example, arbaclofen (a GABA-B agonist) in autism and fragile X – with some anecdotal benefits but inconsistent trial results.

• Neurotransmitter modulators: Besides E-I balance drugs, researchers have tried medications affecting serotonin, dopamine, and other neurotransmitters implicated in ASD. For instance, some studies tried selective serotonin reuptake inhibitors (SSRIs) to reduce repetitive behaviors in autism, but results have been inconclusive (and in younger children, SSRIs often caused side effects without clear benefit on repetitive behaviors). Given the overlap of autism with ADHD symptoms, non-stimulant ADHD meds like guanfacine (an alpha-2 agonist) have shown efficacy for hyperactivity in ASD as well.

• Anti-inflammatory and metabolic treatments: Building on the immunologic findings, trials of anti-inflammatory agents (e.g. broader use of ibuprofen, or cytokine blockers) are mostly in experimental stages or animal models. One small trial tested a fever-reducing medication (sulforaphane, from broccoli sprout extract) after noticing that some autistic individuals temporarily improve during a fever (possibly due to heat-shock proteins); that trial reported some behavioral improvements, but larger studies are needed. On the metabolic front, given observations of oxidative stress and mitochondrial dysfunction in some with ASD, supplements like carnitine, coenzyme Q10, or antioxidants have been tried, but rigorous evidence is limited. The folinic acid trial we discussed earlier is a notable success in targeting a metabolic pathway.

• Targeted treatments for genetic subtypes: Perhaps the most promising pharmacological research is happening in defined genetic syndromes that overlap with autism. For example, Fragile X syndrome (the most common inherited intellectual disability, often co-diagnosed with ASD) has been a testing ground for drugs that reduce excessive synaptic protein synthesis – the so-called mGluR5 antagonists were tried in Fragile X patients to improve cognition and autism symptoms. Those particular drugs did not meet trial endpoints, but the research yielded insights. Rett syndrome (a rare X-linked condition where virtually all patients have autism features along with other neurological issues) recently had a breakthrough: trofinetide, a drug based on IGF-1 growth factor, was approved in 2023 as the first-ever treatment for Rett syndrome . Trofinetide led to modest improvements in Rett patients’ communication and motor function. While Rett is distinct from idiopathic autism, this success shows that disease-specific therapies can be developed when a clear biological target is known. Similarly, in Tuberous Sclerosis (where many have ASD), existing drugs like everolimus (an mTOR inhibitor) are being repurposed to see if they can help autism symptoms by correcting the underlying cell growth pathway overactivity. There’s also excitement about genetic therapies: for conditions like Rett or Angelman syndrome (another single-gene disorder with autism features), researchers are working on gene therapies or antisense oligonucleotides to correct the genetic defect. These are in early trials but could be transformative – for instance, a gene therapy that delivers a functional copy of MECP2 (the Rett gene) or turns on the silenced copy of UBE3A (the Angelman gene) might essentially cure those conditions. While this applies to a minority of autism cases, the knowledge gained often reverberates to more common forms of autism, because many of these genes affect shared brain pathways.

In summary, pharmacological treatment in ASD is moving toward more tailored approaches. Right now, medications are mainly used to alleviate associated symptoms (behavioral outbursts, inattention, anxiety, etc.) , which can greatly help individuals participate in school or therapy. The frontier, however, is developing drugs or biologicals that address autism’s core social/communication challenges or that reverse specific developmental impairments – a very tall order due to autism’s heterogeneity. Each year, a few high-quality trials are reported, and though many have disappointing results, even “failures” teach researchers which mechanisms are or aren’t significant. The involvement of major institutions (NIH has an Autism Center of Excellence network, and private foundations like Simons are funding novel drug research) ensures that progress is ongoing. It’s reasonable to expect that in the next decade, we’ll see a handful of new medications or biological therapies approved that target particular subgroups of ASD or specific symptom domains.

Biotechnological and Novel Interventions

Beyond conventional therapies and drugs, a variety of innovative, technology-driven approaches are being explored in autism intervention. Some of these blur the lines between biology and technology, and many are still experimental but show intriguing possibilities:

• Neurostimulation and neuromodulation: Techniques like Transcranial Magnetic Stimulation (TMS) and transcranial Direct Current Stimulation (tDCS) allow non-invasive modulation of brain activity. Researchers have trialed these in ASD to see if stimulating certain brain regions can improve social cognition or reduce repetitive behaviors. For example, low-frequency TMS to the dorsolateral prefrontal cortex has been studied for reducing irritability and repetitive movements, while high-frequency TMS to brain regions involved in social processing has been tried to enhance social functioning. Some small studies reported improvements in executive functioning or language, but results are preliminary. Similarly, neurofeedback (where individuals learn to modulate their own brainwaves with real-time EEG feedback) has been tested to improve attention or relaxation in ASD. While neuromodulation isn’t standard care, it represents a biotech frontier: the idea that we might directly tune brain networks that are under- or over-active in autism. Larger sham-controlled trials are needed, as placebo effects can be strong with such interventions.

• Assistive and augmentative technology: A very practical use of technology is in communication aids. Many non-speaking or minimally verbal autistic individuals benefit from augmentative and alternative communication (AAC) devices. These range from simple picture-exchange systems to sophisticated speech-generating apps on tablets (for instance, an iPad with software where the user taps symbols or letters and the device speaks for them). These technologies aren’t “curing” autism, but they can be life-changing by giving someone a voice and a way to express needs, thoughts, and feelings. Technological advances have made AAC more affordable and customizable. Additionally, wearable devices and smart sensors are being developed to help with safety and monitoring (such as GPS trackers for children who might wander, or biometric sensors that alert to anxiety levels and can cue the person to use calming strategies).

• Virtual reality (VR) and simulations: Some programs use virtual environments to teach social skills or job skills to autistic teens and adults. For example, a VR program might simulate a job interview or a grocery shopping trip, allowing individuals to practice and learn in a low-stress virtual setting before doing it in real life. Early studies indicate this can help with anxiety and skill generalization.

• Artificial intelligence (AI) and machine learning: AI is being harnessed in a couple of ways. One is earlier diagnosis – machine learning algorithms can analyze home videos of infants or use eye-tracking data to flag subtle signs of autism risk, potentially aiding earlier detection. Another is personalized intervention: AI-driven apps can adapt their teaching or therapy strategies to the user’s responses (creating a sort of virtual therapist or coach). Robots with AI are also being used in some autism programs – many children with ASD show interest in robots, which can deliver lessons or social scenarios in a predictable way, thus serving as a bridge to human interaction.

• Microbiome-based interventions: There is growing interest in the gut-brain axis – the bidirectional communication between the gastrointestinal tract and the nervous system. Some autistic individuals have significant gastrointestinal issues, and studies have found differences in their gut microbiome composition. A few small trials and case studies have tested fecal microbiota transplantation (FMT) (essentially, transferring gut bacteria from a healthy donor to the patient) in children with ASD. Remarkably, one open-label study found not only GI improvements but also some improvements in autism symptoms that seemed to persist for two years post-treatment. This created a buzz, but it was uncontrolled and small. The placebo effect and normal development over time can’t be ruled out. Larger controlled trials of FMT and probiotics (beneficial bacteria) are underway. If a gut-brain link is confirmed, it could open a whole new avenue for treatment. At present, though, microbiome therapies for autism are experimental, and families should approach them cautiously (preferably as part of research studies) given unknowns about long-term safety.

• Precision medicine and genetics: A significant “biotech” push in autism is to bring the power of genetics and molecular biology to individualize care. Initiatives like the SPARK project (Simons Foundation Powering Autism Research for Knowledge) are collecting DNA and detailed information from tens of thousands of autistic individuals and their families. The goal is to identify specific genetic subgroups and eventually match them to targeted treatments or specific supports. For instance, if a child has a mutation in a gene affecting neuron excitability, down the line there might be a specific drug to address that. Or if genetic testing reveals a known syndrome like Fragile X or ADNP syndrome, clinicians can anticipate associated medical issues and intervene early (for example, monitoring and treating epilepsy, which co-occurs in many genetic forms of autism). In the future, gene therapy could be considered for certain forms of ASD – this is already on the horizon for Rett syndrome and Angelman syndrome, as mentioned. The technical and ethical challenges are significant, but rapid advances in gene editing (like CRISPR) and gene delivery mean it’s not a far-fetched idea that some children with autism might receive genetic fixes or biologics that substantially alter their developmental course.

All these emerging interventions share a theme: understanding and addressing the biology of autism at a more fundamental level. Whether it’s using brain stimulation to correct neural circuitry, using microbes to adjust metabolism and immunity, or using genetic insight to guide therapies, the trend is toward tailored interventions that go beyond the behavioral surface of ASD. It’s a recognition that autism, for all its behavioral definition, is rooted in biology – and thus biological and technological tools can play a role alongside education and therapy.

It should be noted that with innovation comes the need for caution. For every promising new therapy, there may be hype or premature use. Families sometimes feel pressure to try unproven treatments (like stem cell infusions offered by unregulated clinics, or experimental diets) – these can be ineffective at best or harmful at worst. The involvement of reputable institutions (e.g. NIH funding trials, academic centers running studies) is a good sign that a line of investigation is legitimate. Always, data is king: the most promising interventions will ultimately be those that demonstrate clear benefits in well-controlled studies, and those are the ones likely to become widely adopted.

Ongoing Controversies and Conclusion

Autism research and advocacy have come a long way, yet controversies persist, often in the realm of what (or who) is to blame for autism and how best to address it. As we have seen, many early fears – like blaming vaccines – have been thoroughly debunked by data . Other newer claims (Tylenol, for example) are still debated but lack consensus support . It’s crucial to approach any autism causation claim or proposed treatment with a critical eye. The difference between speculation and evidence lies in the strength of research: a small study or single paper is rarely definitive. True causes or effective treatments tend to gain support from multiple independent studies, large sample sizes, and ideally randomized trials (for interventions).

The current scientific consensus can be summarized as follows: ASD is a multifactorial neurodevelopmental condition resulting from the interplay of many genes and environmental influences. Genetic predisposition is the dominant factor, with dozens of genes known to contribute and likely hundreds more to be discovered  . Environmental factors (from parental age to certain prenatal exposures) also play a role, but usually as modifiers of risk rather than primary causes on their own . Immunological factors (like maternal infection or autoimmunity) are an area of active research and appear to explain some portion of cases, reinforcing that autism’s etiology spans both brain-intrinsic and systemic influences. What has become clear is that there is no single cause of “the autism” – in fact, autism may be best thought of as a collection of similar neurodevelopmental profiles with many different underlying causes. For one child, a rare gene mutation might be the key factor; for another, a combination of common genes inherited from parents plus an environmental trigger might be responsible. This complexity is why pinpointing “the” cause of autism has been so challenging and at times “progress has been remarkably slow and difficult” in this area .

In light of this, controversies often arise when people search for a simple answer to a complicated question. Claims that a single drug, nutrient, or event caused autism in a broad sense are likely oversimplifications. The data show, for example, that autism rates have increased over the past few decades mainly due to better awareness and expanded diagnostic criteria, not an unchecked environmental toxin epidemic  . While one should remain open to new evidence (science can always evolve with new discoveries), extraordinary claims require extraordinary evidence. So far, the most robust evidence continues to point to autism as a neurodevelopmental difference shaped predominantly by genetics, with many contributing factors and modifiers in play.

On the topic of interventions, another kind of controversy sometimes emerges: the divide between aiming to “cure” or significantly alter autism versus accepting and supporting autistic individuals as they are. The neurodiversity movement, which includes many autistic self-advocates, emphasizes that autism is a natural variation of the human condition and calls for supports and accommodations rather than cures. This perspective coexists with the clinical approach that seeks to alleviate disabling aspects of ASD. In practice, what most families and individuals desire is improved quality of life – and that is the goal of any intervention, be it teaching a child to communicate their needs, or finding a medication to help with anxiety, or even a future gene therapy to ease severe intellectual disability. Balancing the push for breakthroughs with respect for autistic persons’ individuality is an ongoing societal conversation.

From a research standpoint, the future is hopeful. Large-scale collaborations (such as the NIH-funded Autism Centers of Excellence and the Simons Foundation’s initiatives) are accelerating discoveries. For instance, genetic studies with tens of thousands of participants are identifying new genetic variants every year, including moderate-effect genes that help explain additional cases of autism  . These findings not only clarify causation but also point to biological pathways that could be targeted by treatments. At the same time, behavioral science is refining how we deliver interventions – making them more accessible, more tailored, and more effective. Even something as simple as early diagnosis is improving with novel tools, which is crucial since early intervention yields the best outcomes .

In conclusion, autism’s causes are multi-dimensional, involving genes, environment, and immune factors in a complex interplay. The most well-supported causes are genetic variants and a few environmental factors like prenatal insults, whereas many rumored causes (like vaccines) have been ruled out  or remain unproven. As for interventions, the most impactful solutions today are behavioral and educational supports that help individuals learn, grow, and participate more fully in life. On the horizon, pharmacological and biotech breakthroughs may address some core challenges of ASD, especially for defined subgroups, thanks to our expanding biological knowledge. Families and practitioners should stay informed through reputable sources (e.g. the CDC, NIH, or the Simons Foundation’s publications) to discern which new findings are grounded in solid evidence.

Autism is often described as a puzzle – and indeed, researchers are steadily filling in the pieces of that puzzle. Each piece, whether it’s a gene discovery or a trial of a new therapy, contributes to a clearer picture of ASD. It’s unlikely that there will be a single eureka moment that “solves” autism. Rather, progress will come from many incremental advances across genetics, neuroscience, psychology, and education. The end goal is a world where we understand ASD well enough to provide each autistic person with the supports or treatments they need to live a healthy, fulfilling life. With the momentum of current science and the dedication of countless families and professionals, that vision is closer to reality than ever before.

Sources: High-quality sources have informed this review, including research summaries from NIH/NIMH  , findings from the Simons Foundation/SFARI , and numerous peer-reviewed studies as cited throughout. We have drawn on large-scale epidemiological studies, genetic research, and clinical trials to distinguish well-supported evidence from speculation, as referenced in the citations above.

That has changed for me. I’m now 98% ChatGPT when I open an AI directly.

DeepSeek gathering dust.

There are about 10,000x more pictures of you, minutes of video, words you have written, recording minutes of your speaking than your grandparents. More than enough to clone you.

Like Hari Seldon from Foundation.

https://en.wikipedia.org/wiki/Hari_Seldon

What the evidence shows:

Sunlight triggers production of vitamin D, nitric oxide, and immunoregulatory lipids—each associated with lower risks of autoimmune disease, depression, metabolic dysfunction, cardiovascular disease, and some cancers. Circadian regulation by natural light profoundly affects sleep, mood, and hormone balance. Outdoor time correlates with improved mental health, reduced inflammation, and even reduced all-cause mortality in some cohort studies.

Skin tone matters:

People with darker skin produce far less vitamin D from the same amount of sunlight, and face higher risks of deficiency-related illnesses when living in low-UV environments (e.g. high latitudes, urban indoor life). For them, frequent, longer-duration sun exposure may be protective, not harmful. People with lighter skin are at higher risk of UV damage and skin cancers; sunscreen makes sense—but total UV avoidance may increase other health risks.

Rethinking the “no safe exposure” mantra:

The dermatological consensus prioritizes skin cancer prevention, often at the expense of broader systemic health. But emerging evidence suggests that zero UV exposure may itself be harmful, particularly in darker-skinned populations. A nuanced, personalized approach—factoring in skin tone, geography, lifestyle, and disease risk—may be more aligned with evolutionary biology and public health.

Introduction to UV light

Ultraviolet (UV) light – spanning wavelengths from 100–400 nm – has long been recognized as a powerful biological agent. Its spectrum is conventionally divided into UVC (100–280 nm), UVB (280–315 nm) and UVA (315–400 nm), with “far-UVC” referring to the short-wavelength portion of UVC (~200–230 nm) now under intense study. UV phototherapy exploits these bands for medical benefit: UVB and UVA phototherapy are established treatments in dermatology, while UVC/far-UVC are emerging for disinfection and potentially novel therapies.  Over the past century UV has found uses far beyond simple sunlamp tanning.  It mediates complex systemic effects – from vitamin D production and blood-pressure modulation to immune regulation via lipid messengers – and is being investigated for diverse conditions (autoimmune diseases, cancers, infections, etc.).  This report reviews the history, mechanisms, clinical applications, and future prospects of UV therapies (UVA, UVB, narrowband UVB, UVC/far-UVC) for skin and systemic diseases, with emphasis on recent research.

Historical Milestones in UV Therapy

Ancient healers recognized sunlight’s healing power: Egyptian and Indian physicians used sun-exposure (heliotherapy) for skin maladies and “leucoderma” (vitiligo) thousands of years ago .  Modern phototherapy began circa 1900 with Niels Ryberg Finsen, who built an arc lamp (“Finsen lamp”) to treat lupus vulgaris (cutaneous tuberculosis) and won the 1903 Nobel Prize for this “light therapy” .  The 20th century saw rapid advances: in 1974 the psoralen+UVA “PUVA” photochemotherapy protocol was introduced for psoriasis .  Subsequently, narrowband UVB (around 311 nm) and high-intensity UVB units were adopted in the 1980s–90s, improving efficacy and safety.  Extracorporeal photopheresis (treating blood with UVA+psoralen) appeared in 1987 for cutaneous T-cell lymphoma and graft-versus-host disease .  Topical photodynamic therapy (visible light plus photosensitizers) became popular in the 2000s for skin cancer.  Meanwhile, UVC germicidal lamps were used since mid-century to sterilize water and surfaces.  In recent years far-UVC (200–230 nm) has emerged for safe disinfection, and basic research (e.g. Byrne et al. 2023) is uncovering how UV induces systemic immune effects via novel lipid mediators .

A timeline of key UV therapy milestones is shown below:

Year/Decade

Milestone

Notes/Citations

∼1500 BC

Ancient heliotherapy

Egyptian/Indian physicians used sunlight for skin “leucoderma” (vitiligo) .

1895–1903

Finsen’s lamp therapy

Niels Finsen treated lupus vulgaris (cutaneous TB) with UV “chemical rays” (Nobel 1903).

1920s–30s

Early UVB phototherapy

Broad-spectrum UVB cabins used for skin disorders (e.g. psoriasis, rickets) – precursor to modern UVB therapy.

1974

PUVA introduced

Oral/topical psoralen + UVA for psoriasis/vitiligo .

1980s

Narrowband UVB; high-intensity lamps

Development of 311-nm UVB bulbs; wider use of targeted lamps (excimer UVB) .

1987

Extracorporeal photopheresis

UVA+psoralen treatment of peripheral T-cells (for CTCL) began .

1990s

UVA1 phototherapy

Longer-wavelength UVA (340–400 nm) units for deep dermal diseases (e.g. atopic dermatitis, scleroderma).

2000s

Photodynamic therapy (PDT)

Topical ALA/PpIX plus visible/UVA light for skin cancer; LED UV devices introduced.

2010s–2020s

Far-UVC disinfection

207–222 nm excimer lamps developed to kill pathogens safely in occupied spaces .

2020s

UV immunobiology discoveries

Systemic effects via skin-nitric oxide, lipid mediators (Byrne 2023), and type I interferon induction (MS studies) .

Biochemical and Immune Mechanisms

UV light triggers diverse biological effects, many beneficial at controlled doses.  Key mechanisms include:

Vitamin D synthesis (UVB 280–315 nm): UVB converts 7-dehydrocholesterol in skin to vitamin D3, later activated to 1,25(OH)_2D (calcitriol). Vitamin D supports bone health, modulates immunity (innate antimicrobial peptides, regulatory T cells) and its deficiency is linked to infections, autoimmune and metabolic diseases. (Note: vitamin D itself is often credited, but some UVB effects may occur independently.) DNA photodamage and immunomodulation: UVB causes cyclobutane pyrimidine dimers (“sunburn DNA” lesions) and transforms trans-urocanic acid to the cis isomer in skin . Both are well-established local immunosuppressive signals .  Briefly, DNA damage in antigen-presenting cells upregulates IL-10 and regulatory pathways, while cis-urocanic acid has immunosuppressive effects in lymph nodes. These contribute to alleviation of T-cell–mediated skin inflammation (e.g. psoriasis) but also underlie photocarcinogenesis risk. Cytokines and Neuroendocrine Signals: UV exposure induces anti-inflammatory cytokines (IL-10, TGF-β) and regulatory T cells (UV-induced “UV-Tregs” and Bregs) via complex skin-immune interactions . UVA exposure also releases nitric oxide (NO) from dermal nitrate/nitrite stores, causing vasodilation and systemic blood-pressure lowering . Sunlight additionally elevates endorphins and serotonin and regulates melatonin rhythms (via retina exposure), with broad effects on mood and circadian health . Lipid Mediators (Scott Byrne et al. 2023): Recent research has uncovered that UV irradiation generates bioactive lipids in skin and draining lymph nodes. Byrne’s group identified six UV-induced lipids (including platelet-activating factor analogs) in lymph nodes that, when isolated, suppressed T-cell proliferation . This suggests UV can induce a “lipidome” of immunoregulatory molecules systemically, contributing to UV’s immune-suppressive and anti-autoimmune effects. UV also raises sphingosine-1-phosphate (S1P) in lymph nodes, retaining T cells and preventing their egress . In short, UV triggers a cascade of signals (DNA damage, hormones, lipids) that modulate both local and systemic immunity. Antimicrobial effects: UVC (particularly 254 nm) and far-UVC (200–230 nm) are directly germicidal, causing DNA/RNA damage in microbes. UVA (320–400 nm) penetrates deeper but can also generate reactive oxygen species and inactivate certain pathogens (e.g. UVA + riboflavin used for blood sterilization). Clinically, new devices deliver UV (e.g. UVA) internally to kill viruses/bacteria in situ (see “Infectious Diseases” below).

Thus, UV’s “therapeutic window” arises from leveraging its immune-modulatory and anti-proliferative effects while limiting DNA damage.  UVA (315–400 nm) penetrates deeply with minimal direct DNA mutagenesis, releasing NO and inducing T-cell apoptosis, whereas UVB (280–315 nm) is more superficial but potent for vitamin D and keratinocyte modulation. Far-UVC (~222 nm) efficiently inactivates pathogens with negligible skin penetration . The table below summarizes UV bands:

UV Band

Wavelength (nm)

Penetration/Effects

Clinical Uses

Risks

UVA (320–400)

320–400

Deep dermal penetration; ROS generation; immunosuppressive; NO release .

PUVA/Psoralen+UVA: psoriasis, vitiligo, CTCL; UVA1 (340–400) for atopic dermatitis, scleroderma; photopheresis.

Photoaging, melanoma risk, cataract (if unprotected).

UVB (280–315)

280–315

Epidermal action; DNA photolesions; vitamin D synthesis; strong immunosuppression in skin.

Narrowband UVB (311 nm): psoriasis, eczema, vitiligo, lichen planus, prurigo, CTCL, GVHD .

Sunburn, photoaging, non-melanoma skin cancers (risk ↑ with cumulative dose).

UVC (>280)

100–280 (254 nm)

Germicidal DNA/RNA damage (no tissue penetration through ozone layer).

Disinfection of air, water, surfaces (unoccupied spaces); extracorporeal blood sterilization (UVB/Pyrvinium).

Highly carcinogenic to skin/eyes (used only in controlled unoccupied settings) .

Far-UVC

200–230

Very shallow penetration (<5 μm); kills microbes without harming viable cells .

Emerging for continuous room-air and surface disinfection (hospitals, public spaces) .

Early evidence suggests minimal skin/eye risk (cannot reach nuclei) at low doses .

Dermatologic Applications

Phototherapy is a mainstay in dermatology. UV treatments are indicated for many skin conditions by dampening inflammation, inducing apoptosis in pathologic cells, and altering cytokine profiles.  Common indications include:

Psoriasis: The most classic use. Both narrowband UVB (NB-UVB) and PUVA (psoralen+UVA) are highly effective. NB-UVB (311 nm) induces T-cell apoptosis and slows keratinocyte proliferation; it clears or improves psoriasis in most patients after ~12–24 sessions .  PUVA is even more potent; about 90% of patients achieve clearance with PUVA (90% response rate) .  (However, PUVA is more complex and long-term carcinogenic risk is higher than NB-UVB.) Eczema (Atopic Dermatitis): Moderate-to-severe atopic dermatitis can respond to phototherapy.  Both NB-UVB and medium-dose UVA1 (340–400 nm) are used. UVA1 penetrates deeply, inducing apoptosis of skin-infiltrating Th2 lymphocytes and mast cells . UVA1 and NB-UVB have become standard second-line therapies when topical treatments fail .  Phototherapy typically yields substantial itch relief and skin clearing in refractory eczema. Vitiligo: Repigmenting therapy. NB-UVB is first-line for widespread vitiligo (2–3×/week for months).  It stimulates melanocyte proliferation and migration.  PUVA also works but is more cumbersome.  Often combined with topical agents (steroids, calcineurin inhibitors).  Case studies report significant repigmentation in most patients. Cutaneous T-cell lymphoma (CTCL, e.g. mycosis fungoides): Low-grade skin lymphoma responds to phototherapy.  NB-UVB is effective in early patch/plaque disease; PUVA or UVA1 may be used for thicker lesions.  Extracorporeal photopheresis (blood treated with UVA/8-MOP) is approved for Sézary syndrome and refractory CTCL. Prurigo Nodularis and Chronic Pruritus: UVB phototherapy reduces chronic itch by modulating nerve endings and cytokines. NB-UVB is often used successfully for prurigo nodularis and generalized pruritus . Photodermatoses: Polymorphic light eruption (sun allergy) can be prevented by “hardening” phototherapy in spring (e.g. low-dose UVA/PUVA) to induce tolerance.  Also used for chronic actinic dermatitis. Sclerotic Skin Diseases: High-dose UVA1 is used for localized scleroderma (morphea), systemic sclerosis skin involvement, keloids and scars .  UVA1 induces collagenases in fibroblasts to soften fibrosis . Others: Psoriasis variants (palmoplantar, scalp), lichen planus, erythroderma, granuloma annulare, pityriasis lichenoides – all may improve with appropriate UV regimen.  Actinic keratoses can be cleared by UV photodynamic therapy (ALA+UV).  UVA2 (315–340 nm) or UVB targets can treat mild viral warts (with or without 5-FU).

A comprehensive list of indications is long (see Table below).  In practice, most treatment protocols start with NB-UVB thrice weekly (5–10 minutes) and escalate dose slowly.  UVA1 is given less frequently (up to 30–60 minutes) in high-dose regimens.  Protection of eyes and non-involved skin, and spacing sessions, mitigate adverse effects.

Dermatologic Condition

UV Therapy

Notes/Citation

Psoriasis vulgaris

NB-UVB 311 nm (first-line) or PUVA

Cleared in most patients after 12–24 tx .

Atopic dermatitis

NB-UVB or UVA1

UVA1 induces T-cell apoptosis in eczema .

Vitiligo

NB-UVB, PUVA

Stimulates melanogenesis; requires months of tx.

Cutaneous T-cell lymphoma (MF)

NB-UVB, PUVA, ECP

Photopheresis approved for Sézary/CTCL.

Prurigo nodularis / chronic itch

NB-UVB

Reduces neural inflammation .

Scleroderma/morphea

High-dose UVA1

Softens collagen via MMP induction .

Polymorphic light eruption

Hardening course (UVA/PUVA)

Seasonal prophylaxis.

Lichen planus

NB-UVB or PUVA

Lesions often improve.

Granuloma annulare

NB-UVB

Topical therapy adjunct.

Pruritic dermatoses (eg. urticaria pigmentosa)

NB-UVB or UVA1

Melanocytosis often ameliorated.

Table: Selected dermatologic indications for UV phototherapy. (Sources: DermNet NZ and review articles .)

Systemic and Autoimmune Applications

Beyond skin, UV light affects systemic immunity and has been investigated (mostly experimentally) in several non-dermatologic conditions:

Multiple Sclerosis (MS): Epidemiology and trials suggest sunlight is protective. High latitude (low sun) is a strong MS risk factor. A 2020 study found that MS patients with greater sun exposure and higher vitamin D levels had milder disease and fewer MRI lesions . Sunlight directly induces type I interferon–related gene signatures in immune cells, which may explain UV’s benefit in MS independent of vitamin D . Ongoing trials of whole-body UVB phototherapy in MS aim to exploit this effect. Rheumatoid Arthritis (RA): Large cohorts (Nurses’ Health Study) reported that women in high-UVB regions had a ~20% lower incidence of RA than those in low-sun areas . UVB increases vitamin D and may downregulate Th1/Th17 pathways implicated in RA. Some clinics in Asia use NB-UVB or sun exposure as adjunctive therapy for active RA, but formal phototherapy for RA is not established. Inflammatory Bowel Disease (IBD): IBD (Crohn’s, ulcerative colitis) incidence shows a north–south gradient (higher in low-UV regions) .  In animal models, UVB exposure ameliorated colitis, partly via vitamin D and regulatory T cells. An ongoing clinical trial (Australia) is testing low-level laser (photobiomodulation) therapy in Crohn’s disease.  While still experimental, UV-driven reduction of gut inflammation is hypothesized through both vitamin D–dependent and independent immune pathways . Type I Diabetes, Lupus, and Other Autoimmunity: Observationally, low sun exposure is linked to higher rates of type 1 diabetes and lupus (paradoxically, lupus flares with sun though). The “latitude-dependent autoimmune disease” hypothesis posits that UVR exposure downregulates autoimmunity for conditions such as MS, RA, type-1 diabetes, Crohn’s, SLE, etc . Experimental UVB or UVA treatment of mice predisposed to autoimmune diabetes reduced disease incidence (vitamin D–dependent and independent effects). Human clinical use of UV in these diseases is not standard, but vitamin D supplementation and controlled sun exposure are sometimes advocated as preventive measures. Cancer (non-skin): Systemic benefits of sun on internal cancers have been noted. For example, higher sun exposure is associated with lower colon, breast, prostate cancer rates, possibly via vitamin D and immune effects . Clinical trials are exploring UVA phototherapy for certain cancers: e.g. isolated limb photochemotherapy (Illuminox device) uses UVA to palliate melanoma metastases. Photodynamic therapy (although often visible light) is increasingly used for early lung and esophageal cancers via endoscopic drug activation. Bone and Metabolic Health: UVB–vitamin D is crucial for bone mineralization; deficiency causes osteomalacia/rickets. Moreover, vitamin D insufficiency has been linked to metabolic syndrome and type 2 diabetes. A 2023 review noted sunlight deficiency contributes to metabolic disorders and childhood myopia . Thus, moderate UV exposure may have broad public health implications for endocrine and metabolic balance.

In summary, sunlight’s systemic effects – from vitamin D to interferons to lipids – suggest potential for phototherapy beyond dermatology. To date, clinical UV treatments are mostly dermatologic, but research continues into uses like MS modulation, gut inflammation, and even psychiatric/mood disorders.

Infectious Disease and UV

UV light’s germicidal properties have long been used for sterilization.  In medicine:

Environmental Disinfection: UVC (254 nm) lamps are widely used to disinfect hospitals, labs, air and water supplies.  Far-UVC (200–230 nm) is an exciting new tool: at these wavelengths, UVC inactivates bacteria and viruses as effectively as 254 nm but cannot penetrate the outer dead cell layers of human skin or the tear film of eyes .  Studies show far-UVC (222 nm) delivered at low intensity can inactivate >99.9% of airborne coronaviruses (OC43, 229E) in minutes .  Continuous far-UVC fixtures (e.g. Ushio’s “Care222” lamps, UV Angel’s Air+) are now commercialized to sterilize occupied rooms (ORs, classrooms) safely. Therapeutic Phototherapy for Infections: There are intriguing, early-stage clinical uses of UV in vivo against pathogens. For example, narrowband UVA (around 340–360 nm) was delivered via a novel endotracheal tube device in ventilated COVID-19 patients .  In a first-in-human trial, 5 critically ill patients received 20-minute intratracheal UVA sessions for 5 days; results showed a significant drop in SARS-CoV-2 viral load in airway samples . (No adverse effects were noted.) Similarly, UVA (with photosensitizer) is used to disinfect blood products (Terumo’s Mirasol system, using riboflavin+UV to inactivate viruses/bacteria in platelets/plasma). Photodynamic therapy (visible light + photosensitizer) is used for localized infections (e.g. PDT for chronic sinusitis or diabetic foot ulcers with photosensitizers). Antimicrobial Skin Effects: UVB phototherapy boosts local antimicrobial defense in skin; chronic UV exposure upregulates cathelicidin and other peptides. Some clinics use UVB to treat chronic wound infections, although evidence is limited. In the past, Finsen’s lupus therapy was essentially phototherapy for a TB infection.

While UVC phototherapy on living tissue is largely avoided due to carcinogenicity, far-UVC and UVA methods are opening new frontiers against pathogens without damaging human cells.  These disinfection applications are especially promising in hospital and public health settings.

Regional and Global Perspectives

Adoption of UV therapies varies by region. In Europe and North America, dermatologists commonly use NB-UVB and PUVA for skin diseases; UVA1 units are widely available in European centers (e.g. Germany, Austria) but are rarer in the U.S. due to cost and space constraints . In Asia, phototherapy is also well established – for example, Japanese clinicians report success using UVB for early cutaneous lymphomas . Some regions combine UV with climatic therapy: Middle Eastern “Dead Sea climatotherapy” uses intense natural UV and mineral-rich water for psoriasis.

Cultural factors also influence general UV exposure: sunscreen use is higher in Australia and parts of Europe, whereas some Asian countries have traditionally favored umbrella-shading to avoid tanning (affecting baseline sun doses). These lifestyle patterns correlate with disease incidence: high-latitude/northern countries have more MS, Crohn’s, RA (low sun regions) compared to sunny equatorial areas .  Notably, an international consensus concluded that broad-spectrum sunscreen use does not typically cause vitamin D deficiency in healthy populations – real-world usage allows some UV through, and supplements mitigate risk. However, patients with photosensitive disorders (e.g. lupus, xeroderma) who use rigorous photoprotection do require vitamin D monitoring and supplementation .

Sunscreen, Indoor Lifestyles, and Public Health

Modern lifestyles have raised concerns about “sunlight deficiency.”  Many people spend >90% of time indoors and evenings under artificial light.  Excessive screen time appears linked to lower sun exposure: one study found US children with ≥5 hours daily recreational screen time had more than double the odds of vitamin D deficiency compared to those with <2 hours . By contrast, outdoor play was protective. Urbanization and office jobs further limit UVB exposure, potentially exacerbating vit D insufficiency and immune-related disorders.

The role of sunscreen is double-edged. Sunscreens effectively block erythemal UVB (SPF15 blocks ~93%, SPF50 ~98% ), which could theoretically reduce cutaneous vitamin D production and other benefits. However, expert panels find that normal sunscreen use does not generally cause widespread vitamin D deficiency . In trials where sunscreen was applied perfectly, vitamin D synthesis is reduced, but real-world use is imperfect. Thus public health advice remains to use sun protection to prevent skin cancer, supplemented by diet or low-dose sunlight if necessary.

On the other hand, some UV effects (nitric oxide release, modulation of circadian hormones) require UVA exposure. Overuse of broad-spectrum sunscreens could blunt these. The optimal balance between photoprotection and systemic benefits is an area of active debate. Some researchers even posit a “darkness pandemic,” arguing that lack of sunlight contributes to chronic diseases (metabolic syndrome, some cancers, myopia, mood disorders) . For example, sunlight helps regulate melatonin and mood, and is essential to prevent childhood myopia . While evidence is still emerging, reasonable unprotected sun exposure (e.g. 10–15 minutes a day) may be advocated for health, followed by protection once the skin is warm or pink.

Biotech Innovations and Commercial Devices

A number of startups and companies are commercializing UV technologies:

Disinfection/UVC: Uviquity (founded 2025) secured seed funding to develop far-UVC LEDs for continuous pathogen inactivation in occupied spaces . Existing products include Ushio’s “Care222” excimer lamps and fixtures by Xenex and UV Angel, which use far-UVC in hospitals and schools for air/room sterilization. Traditional UVC devices (e.g. Xenex Pulsed Xenon, Sterilray units) disinfect unoccupied rooms and water supplies. Phototherapy Equipment: Established manufacturers (Daavlin, Waldmann, Philips) produce UVB and UVA booths and home units. LED UVB devices (311 nm LEDs) are emerging.  Mobile or small-footprint phototherapy cabins enable home treatment under dermatology supervision, improving access. Photopheresis and Blood Safety: Mallinckrodt’s Therakos system and similar devices use UVA+psoralen to treat blood (for CTCL and transplant rejection).  Terumo’s Mirasol uses UVB and riboflavin to inactivate pathogens in platelets/plasma. Photobiomodulation Devices: Though many use red/near-IR light, some home devices combine UV (e.g. UVA LEDs for nail fungus). Clinical trials are testing UVC ring lamps for MRSA decolonization in patients. Eye and Skin Care: LED-based sunlamps and narrowband devices target vitamin D deficiency or skin conditions. In dermatology, biologic-sensitizing agents (“photoagonists”) are being explored to enhance phototherapy efficacy.

Each year sees new UV-related products. For example, Symbyx (Australia) is investigating laser therapy devices for inflammatory bowel disease (an extension of photobiomodulation) . The trend is toward safer, targeted UV sources (semiconductor far-UVC, filtered lamps) and smart systems that track dose. Notably, policy groups (e.g. IUVA) are issuing guidelines on UV disinfection, reflecting the growing mainstream interest in UV technologies.

Potential and Risks of UV Bands

Far-UVC (200–230 nm):  Far-UVC’s promise lies in pathogen inactivation with minimal harm to humans.  Studies show 222 nm light at low doses kills >99% of airborne viruses in minutes and destroys bacteria efficiently . Importantly, 222 nm cannot penetrate the stratum corneum or tear film, so it causes almost no DNA damage in skin models .  Continuous low-level exposure to 222 nm (within safety limits ~3 mJ/cm²/h) could reduce viral loads ~90% every 10 minutes in a room . The main challenges are engineering (making reliable far-UVC LEDs and fixtures, eye safety over long use) and regulatory approval for occupied use.  If these are overcome, far-UVC could be widely deployed (e.g. in airports) for infection control without irradiating people’s skin.

Traditional UVC (254 nm): The germicidal “medium-pressure” UVC lamp is highly effective against pathogens and is standard for water and surface disinfection. However, it is carcinogenic to skin/eyes, so use is confined to empty spaces or shields (e.g. upper-room UVGI, laminar flow hoods).  Its role in therapy (on patients) is minimal due to safety.

UVB (280–315 nm): The therapeutic “workhorse” for skin diseases. Narrowband UVB (311 nm) offers maximal benefit with reduced carcinogenicity versus broadband.  However, cumulative UVB doses can still increase non-melanoma skin cancer risk.  For example, PUVA (which includes UVA) significantly raises squamous carcinoma risk after 100+ treatments . Thus, photo-oncology safety monitoring (skin exams, limiting tx number) is critical.

UVA (315–400 nm): UVA penetrates deeply, is safer in the short term (no direct DNA photolesions) but induces oxidative stress. Long-term UVA exposure (as from tanning beds or PUVA) is linked to melanoma and aging. New UVA1 therapies (340–400 nm) deliver high fluences quickly and may reduce cumulative risk, but data are limited.  UVA also causes ocular damage (photokeratitis, cataracts) without protection – all UV therapies require eye shields.

In summary, every UV band has pros and cons.  Balancing dose, spectrum, and treatment frequency is vital.  Modern phototherapy protocols minimize risks by using narrowband lamps, limited courses, and patient selection.  The “band gap” approach (e.g. far-UVC for viruses, UVA1 for deep lesions) is key to exploiting UV’s potential safely.

Sunscreen and “Sunlight Deficiency”

Sun protection is critical to prevent UV harm, but questions remain about losing “good” UV. High-SPF broad-spectrum sunscreens block nearly all erythemal UVB , which indeed can impede vitamin D synthesis in theory. However, a 2019 international panel concluded that realistic sunscreen use does not cause widespread vitamin D deficiency in healthy people . In practice, people apply sunscreen imperfectly and get incidental sun. Thus, public health advice remains: use sunscreen to prevent skin cancer, and obtain vitamin D through moderate sun or supplements if needed.

That said, very strict photoprotection (for patients with lupus or photosensitivity) can lower vitamin D, so those individuals should be monitored . Furthermore, sunscreen blocks UVA too, which could blunt UV-mediated NO release or endorphin effects. No clinical trials have tested whether sunscreen use reduces the systemic benefits of incidental sun beyond vitamin D. For now, experts recommend sensible sun exposure (short periods unprotected, then apply sunscreen) to balance benefits and risks .

The rise of indoor lifestyles has raised concerns of “sunlight deficiency” contributing to chronic disease.  Aside from vitamin D, reduced UV exposure may perturb circadian and immune function. For example, insufficient daylight exposure is linked to seasonal affective disorder and myopia in children . The COVID-19 pandemic, by keeping people indoors, also reignited interest in UV for immune health. Some commentators advocate “sunlight as medicine,” suggesting that deliberate low-dose exposure (beyond vitamin D) may have untapped public health benefits . While more research is needed, it seems plausible that endemic diseases of modern life (autoimmunity, metabolic syndrome, depression) could be partly driven by chronic lack of natural UV stimuli.

Future Directions

Research and technology developments point to several promising frontiers in UV therapy:

Far-UVC and Disinfection: Engineering of solid-state far-UVC emitters is progressing.  Uviquity (semiconductor-based) and evolving designs of 222-nm LEDs could soon enable ubiquitous safe-room decontamination . Regulatory agencies are studying long-term safety, but if approved, far-UVC might become standard in public health (e.g. real-time air sterilizers in hospitals and transit). Personalized Photomedicine: Genomic and microbiome profiling may identify subgroups who respond best to UV (e.g. MS patients with certain sun-sensitivity genes). Dosimetry could be individualized. Combining phototherapy with systemic biologic drugs (e.g. psoriasis biologics + UV) is under study to achieve synergy or reduce drug dose. New Indications: Trials are exploring UV/light for novel uses. Besides IBD (see above), trials of UVB for type-2 diabetes (via vitamin D and inflammation), or UVA1 for localized scleroderma are ongoing. NASA research looks at UV for immune support in spaceflight. Photobiomodulation beyond UV (LED red/NIR) is a hot field, and some combine it with UV/blue light. Mechanistic Insights: Scott Byrne’s lipidomics and related immunology studies open doors for targeted modulation. If specific lipid mediators are key, one could conceive “UV mimetics” or topical agents that induce similar pathways without UV exposure. Conversely, understanding UV’s systemic signaling could inspire new drug targets (e.g. S1P modulators). Combination Therapies: Photodynamic therapy (PDT) using UV or visible light with photosensitizers is being extended to internal tumors and antimicrobial treatment. There is interest in using UV to boost vaccine responses (e.g. slight UV dose at vaccination site to enhance antigen presentation). Sunscreen Innovation: New sunscreens that selectively filter harmful wavelengths while permitting “beneficial” UVA (for NO) or visible light are being researched. This could optimize protection vs. wellness.

In conclusion, UV phototherapy remains a vibrant field, blending century-old treatments with cutting-edge science.  Clinicians and researchers now recognize that UV’s effects are not confined to the skin: it is a systemic immunomodulator and environmental factor of human health.  While risks (cancer, photoaging) mandate caution, the expanding repertoire of UV wavelengths and devices allows ever-more-specific interventions.  Ongoing trials and startups (e.g. Uviquity’s far-UVC, Symbyx’s gut-light therapy) indicate that UV-based therapies will continue to grow in scope.

Key takeaways: Sunlight (UV) has both historical and modern medical uses beyond vitamin D.  Phototherapy (UVA/UVB) is proven in dermatology (psoriasis, eczema, vitiligo, etc.) and is under investigation for systemic conditions (MS, IBD, RA).  Far-UVC is an emerging safe disinfection tool.  UV triggers complex mechanisms (vitamin D, nitric oxide, cytokines, lipids) that can suppress autoimmunity and fight pathogens.  Sunscreen mitigates UV risks but modest sun exposure may confer systemic benefits.  Urban lifestyles reducing UV exposure are being scrutinized as possible contributors to chronic disease.  New biotech ventures (e.g. Uviquity, UV Angel) and device innovations promise expanded UV therapies in health care.

Sources: Recent peer-reviewed studies and reviews were used to compile this analysis (e.g. Byrne et al. 2023 , Buonanno et al. 2020 , Hönigsmann 2013 , DermNet NZ references , and others). All factual claims are supported by citations above.

Here’s a concise table benchmarking human information intake and output across different modalities, roughly quantifying bandwidth in terms of bits per second (bps) or words/ideas per minute (wpm/ipm) where applicable. These are estimates grounded in cognitive psychology and neuroscience research.

Modality

Type

Bandwidth Estimate

Notes

——

Listening

Input

~60 bps / ~150 wpm

Conversational speech; limited by working memory.

Talking

Output

~60 bps / ~120–150 wpm

Slower for complex concepts; real-time feedback modulates clarity.

Reading (text)

Input

~300 bps / ~250–300 wpm

Faster than listening; denser information per unit time.

Writing (text)

Output

~20–40 bps / ~20–40 wpm

Very low throughput, but high precision and permanence.

Images (viewing)

Input

~10,000 bps (estimated)

High parallelism; extremely rich but hard to quantify precisely in wpm.

Video (watching)

Input

~20,000–40,000 bps

Combines audio and visual streams; good for narrative and emotion.

Gestures/Body Language

Input/Output

~10–50 bps (wildly context-dependent)

Subtle, low-precision, high-contextual load; rapid emotional signaling.

Touch

Input

~1,000 bps (Braille: ~30–50 wpm)

High fidelity for texture, pain, pressure; rarely used for abstract info.

Smell

Input

~1 bps (symbolically negligible)

Powerful memory triggers, poor for abstract communication.

Taste

Input

<1 bps

Like smell, limited symbolic bandwidth.

Facial Expressions

Output

~10–100 bps

Rapid and universal emotional cues; not good for detailed data.

At that time:

The University of Florida was still segregated. The school admitted its first Black undergraduate student in 1962, but Black athletes did not begin playing varsity football there until 1969. In 1965, the Gators football team was entirely white.

Context:

The Southeastern Conference (SEC), to which Florida belongs, was one of the last major conferences to integrate its athletic programs. Full desegregation of football programs across the South happened slowly through the late 1960s into the early 1970s

Love Trump. Love how he operates. He’s like a king. 

The regional geopolitics changed after MBS took power. Saudi learned in the 2000s that it should not let other countries dominate the regional politics. 

The main players have always known that the US would eventually leave the region. That means Israel, Iran, Turkey. They all adopted aggressive policies.

Qatar and Bahrain, are client states of the US including the UAE. All of them have American bases. The only reason they are independent countries is because they have American bases. Those bases are for America to project its power but also to keep those countries independent. 

Nobody wants a war with Iran. Nobody wants them to have nuclear weapons either. But the policy under Obama and Trump and Biden and Trump again is best if it defuses the conflict with Iran. (that may not be the Israeli view.) 

The problematic mixed strategy from Qatar is actually with full US consent. That means the activities of Qatar to “destabilize the region“, eg funding Hamas. And if it’s with US consent that means it’s also with Israeli consent. Furthermore, meaning Israel has always wanted an action out of Gaza that would justify the full domination of that area. The strategy of Israel is to take control of the West Bank parentheses a project that is well underway) and now that project for Gaza is also underway.

Saudi is now projecting power. Overdue.

The current Trump position is very favorable to Saudi. It supports a I’m gonna go upstairs no US strategy in the region. Saudi is fully capable of dominating its neighbors. Perhaps the exception is Iran, or Iran is strong defensively not offensively.  

Summary of Mundell’s Optimum Currency Area (OCA) Theory Quantitative Criteria and their Application to Europe Assessment of Other Potential Currency Regions Risks, Constraints, and Policy Implications

1. MUNDELL’S OCA THEORY — FRAMEWORK OVERVIEW

Core Idea: A region should adopt a single currency if the economic benefits from eliminating exchange rate volatility outweigh the costs of losing independent monetary policy.

Mundell introduced four key criteria to evaluate this tradeoff:

Criterion

Definition

Proxy Metrics (Quantitative)

1. Labor Mobility

Ability of workers to move across borders to offset asymmetric shocks

Migration flows, unemployment divergence

2. Capital Mobility & Price/Wage Flexibility

Capacity of capital and prices to adjust smoothly

FDI/portfolio flows, price dispersion, wage rigidity indices

3. Fiscal Transfers

Mechanism to redistribute resources across regions

Central budget size, fiscal equalization transfers (% of GDP)

4. Symmetry of Shocks

Extent to which countries experience similar economic fluctuations

GDP correlation coefficients, standard deviation of growth shocks

Mundell argued that if these conditions are met, the cost of foregoing monetary sovereignty is lower than the benefits of currency integration.

2. APPLICATION TO EUROPE — WHY THE EURO?

Mundell’s key insight in applying OCA to Europe was that Europe already had high trade integration and relatively symmetric business cycles among core countries, especially Germany, France, and Benelux.

Quantitative Analysis

a) Trade Integration (Benefit of a Common Currency)

Intra-EU trade/GDP: ~25–35% for most Western EU countries pre-Euro. Currency transaction costs: Estimated to be ~0.4–0.5% of EU GDP pre-Euro. Elimination of exchange rate risk: Increases trade by ~8–16% (Rose Effect, 2000).

b) Symmetry of Shocks

GDP Growth Correlation (1990s, pre-Euro): Germany–France: 0.85 Germany–Netherlands: 0.88 Germany–Italy: 0.70 Asymmetric shock risk was deemed manageable for core countries.

c) Labor and Capital Mobility

Labor Mobility Index (relative to US): ~0.2 (EU), 1.0 (US) Capital Mobility: High post-Maastricht (1993) with unified financial markets.

d) Fiscal Transfer Mechanism

EU Budget: ~1% of EU GDP Contrast: US Federal Budget: ~20% of GDP, with significant stabilizers (e.g. unemployment insurance).

Conclusion:

From a Mundellian view, core Europe (Germany, France, Benelux) closely approximated an OCA. However, Southern and Eastern Europe had lower shock symmetry and labor mobility, making them less ideal candidates (as seen during the Eurozone debt crisis).

3. OTHER REGIONS — WHICH COULD ADOPT A COMMON CURRENCY?

Let’s assess several candidate regions using OCA criteria.

A. North America: US–Canada–Mexico

Criterion

Metric (Canada–US)

Metric (Mexico–US)

Comment

Trade Integration

Canada: ~75% with US

Mexico: ~80% with US

High

GDP Correlation

~0.85 (Canada), ~0.65 (Mexico)

Moderate asymmetry

Labor Mobility

Low across borders

Very low

Weak

Fiscal Transfers

None

None

Weak

Conclusion: Canada and the US come closer to an OCA, but political and sovereignty issues dominate. NAFTA/USMCA increases integration, but insufficient labor mobility and no fiscal transfer system make a shared currency risky.

B. East Africa (EAC: Kenya, Tanzania, Uganda, Rwanda, Burundi)

Metric

Value

Intra-EAC trade

~15–20% of total trade

GDP correlation

~0.60–0.75 (Kenya-Tanzania, Kenya-Uganda)

Labor mobility

Moderate (informal high), formal barriers remain

Fiscal transfers

Very limited (nonexistent)

Conclusion: Moving toward convergence, but low institutional capacity, fiscal heterogeneity, and lack of transfer mechanisms make premature monetary union risky.

C. ASEAN (Southeast Asia)

Country Pair

Trade/GDP

GDP Correlation

Labor Mobility

FX Shock Absorption

Singapore–Malaysia

High

~0.80

High

High

Indonesia–Thailand

Moderate

~0.55

Low

Moderate

Conclusion: High diversity in inflation targeting, fiscal regimes, and political systems makes ASEAN not viable as a currency area currently. However, sub-groups (e.g. Singapore-Malaysia-Brunei) could be evaluated further.

4. RISKS AND STRATEGIC INSIGHTS

Risks of Currency Union Without OCA Compliance:

Inability to devalue currency during shocks (e.g., Greece 2010s). Fiscal rigidity without central stabilization fund. Political fallout from economic divergence.

Strategic Insights (McKinsey-style takeaways):

Insight

Recommendation

Trade Intensity Is Necessary But Not Sufficient

High trade volumes are a prerequisite but don’t offset asymmetric shocks without labor/fiscal flexibility.

Small Economies Benefit Most

Countries with historically volatile currencies or inflation benefit more from the stability of a shared currency (e.g., Baltics joining Euro).

Fiscal Union Must Precede or Parallel Monetary Union

Without a central budget >10% of GDP and automatic stabilizers, currency areas are vulnerable to crises.

Institutional Readiness Matters

Even if economic metrics align, weak governance can derail a common currency’s success.

Conclusion

Mundell’s theory implies that Europe’s core states were uniquely suited for a common currency due to high trade, economic correlation, and institutional alignment. Other regions—North America (Canada-US) or East Africa—may eventually qualify, but only with deeper economic integration and fiscal harmonization. His framework remains the gold standard for evaluating currency unions and highlights the delicate tradeoffs between sovereignty and stability.

OK so did the Euro make sense? Where else would it matter?

The impact of the euro on European growth rates is contested among economists, and research findings vary depending on the methodology, time frame, and countries included. While some studies find modest positive effects, others argue the euro has had little to no significant impact—and in some cases, even negative effects for weaker economies.

Here’s a structured, quantitative analysis like a McKinsey case study would present:

1. WHAT IS THE COUNTERFACTUAL?

The counterfactual is: What would GDP growth have been in euro area countries had they retained national currencies?

Economists estimate this using models such as:

Synthetic Control Method (constructing a doppelgänger country) Difference-in-Differences (DiD) approach Panel regressions with control variables

2. EMPIRICAL FINDINGS FROM MAJOR STUDIES

Study

Methodology

Countries

Time Period

Euro Growth Effect

Gagnon & Hinterschweiger (2013, Peterson Inst.)

Synthetic control

Eurozone core

1999–2012

~0.5–1.0 pp lower GDP per capita vs. synthetic

Campos, Coricelli, Moretti (2019, J. Econ Lit)

Synthetic control (comprehensive)

11 Euro adopters

1993–2013

+5–10% cumulative GDP gain

Eichengreen & Boltho (2008)

Narrative + empirical

France, Italy, Germany

1999–2007

Negligible net effect

Berger & Nitsch (2008, JME)

Gravity model (trade → growth)

EU15

1995–2005

+5–10% increase in trade, but <0.1% GDP growth gain/year

IMF (2017 Euro Area Review)

Panel regression

All EA19

1999–2016

Modest benefit for core, divergence for periphery

Summary:

Core countries (Germany, Netherlands, Austria) saw small to moderate gains, mostly through trade and lower interest rates. Periphery (Greece, Italy, Portugal, Spain) saw short-term booms, then long-term stagnation due to loss of monetary tools and competitiveness. Overall GDP growth differential attributed to the euro: +0.1 to +0.3 percentage points annually for core countries. Net neutral or negative for several periphery countries.

3. STRUCTURAL CHANNELS OF IMPACT

Channel

Positive Impact

Negative Impact

Trade Integration

5–15% increase in intra-EA trade volumes

Trade boost didn’t always translate into productivity or GDP gains

Price Transparency

Improved consumer choice and cross-border competition

Minor effect on macro growth

Interest Rate Convergence

Lower borrowing costs in South

Encouraged credit booms, real estate bubbles

Loss of Exchange Rate Flexibility

Greater price stability

Inability to respond to asymmetric shocks (e.g. 2010s debt crisis)

ECB Credibility

Anchored inflation expectations

“One-size-fits-all” policy hurt diverging economies

4. QUANTITATIVE ILLUSTRATION (EUROZONE VS NON-EURO EU STATES)

Country Group

Avg. Annual GDP Growth (1999–2019)

Std. Dev.

Notes

Eurozone (EA12)

~1.4%

1.2%

Germany ~1.3%, Italy ~0.6%

Non-Euro EU (e.g., Poland, Czech Rep, Sweden)

~3.0%

2.1%

Stronger catch-up growth

However, much of the difference is due to initial income levels and post-transition dynamics, not just currency choice.

5. CONSULTING-STYLE TAKEAWAYS

Insight

Implication

Euro ≠ Growth Engine

Growth impact of euro is marginal compared to structural reforms, demographics, and investment

Competitiveness Trap

Southern Europe lost export competitiveness due to fixed exchange rates

Trade ≠ Productivity

Trade increases from euro didn’t translate to proportional GDP or TFP growth

Institutions Matter More

Convergence depends more on regulatory and fiscal alignment than currency alone

Conclusion

Most rigorous economic studies conclude that the euro modestly improved growth for core economies, had neutral to negative impacts on the periphery, and that macro benefits fall short of initial expectations. The euro’s real value has been institutional and political—cementing European integration—rather than as a growth accelerator.

So if it is only one piece of an equation, what else matters?

To unlock the full economic benefits of a common currency and economic zone like the Eurozone, reforms must address the structural limitations of monetary union without fiscal, financial, labor, and regulatory integration. Think of the euro as a bridge half-built; without completing the full architecture of union, its economic potential is capped.

Below is a strategic, component-by-component breakdown of major reforms—completed, partial, and pending—along with estimated impacts on Eurozone GDP (long-run delta from baseline, in % of GDP), derived from empirical studies, ECB estimates, IMF projections, and consulting-level synthesis.

1. Banking Union (Financial Market Integration)

Goal: Reduce fragmentation in financial markets, ensure that eurozone banks operate under common supervision, and break the bank-sovereign doom loop.

Component

Status

Economic Rationale

Est. GDP Impact

Single Supervisory Mechanism (SSM)

Complete

Common ECB oversight of systemic banks

Already helping

Single Resolution Mechanism (SRM)

Partial

Bank resolution fund too small (~€80B)

+0.3–0.5%

European Deposit Insurance Scheme (EDIS)

Incomplete

Would stabilize banking confidence across borders

+0.5–0.7%

Capital Markets Union (CMU)

Ongoing

Mobilizes private capital across EU borders

+0.2–0.4%

Subtotal GDP Delta (long-run): +1.0–1.6%

2. Fiscal Union & Stabilization Mechanisms

Goal: Allow for centralized counter-cyclical spending and automatic stabilizers to offset asymmetric shocks (like in the U.S.).

Component

Status

Economic Rationale

Est. GDP Impact

Central Fiscal Capacity (e.g., EU budget >1% GDP)

Weak

US federal budget ~20% of GDP vs. EU ~1%

+1.0–2.0%

Eurobond or Common Debt Instrument

Partial (NextGen EU is a start)

Pooling risk lowers borrowing costs for weaker members

+0.3–0.6%

Unemployment Insurance Scheme

Not implemented

Provides automatic transfers during downturns

+0.4–0.7%

Harmonization of Tax Policy

Partial

Reduces distortions and race to the bottom

+0.2–0.4%

Subtotal GDP Delta (long-run): +1.9–3.7%

3. Labor Mobility & Integration

Goal: Increase the ability of workers to move freely and efficiently across borders.

Component

Status

Economic Rationale

Est. GDP Impact

Portability of benefits/pensions

Partial

Disincentives remain for cross-border moves

+0.2%

Recognition of professional qualifications

Partial

Legal and bureaucratic hurdles limit skilled migration

+0.2–0.3%

Language, housing, and education integration

Under-addressed

Cultural barriers block flows unlike U.S.

+0.3–0.5%

EU-wide job matching system (EURES upgrade)

Weak

Labor market frictions remain high

+0.2–0.3%

Subtotal GDP Delta (long-run): +0.9–1.3%

4. Regulatory & Structural Convergence

Goal: Align product, service, and business regulations to increase scale economies and reduce barriers to competition.

Component

Status

Economic Rationale

Est. GDP Impact

Services Directive enforcement

Incomplete

Services = 70% of GDP; still many national barriers

+1.0–1.5%

Digital Single Market

Partial

Potential to unleash cross-border digital growth

+0.5–0.8%

SME and startup ecosystem convergence

Fragmented

Access to capital, regulation varies by country

+0.3–0.5%

Subtotal GDP Delta (long-run): +1.8–2.8%

5. Innovation, Infrastructure, and Education Investment

Goal: Align R&D, digital infrastructure, and education policy to improve total factor productivity (TFP).

Component

Status

Economic Rationale

Est. GDP Impact

Coordinated R&D funding (e.g. Horizon EU)

Ongoing

Improve innovation spillovers

+0.4–0.7%

5G/Digital infrastructure across South/East

Underdeveloped

Closes competitiveness gap

+0.5–0.8%

EU-wide education and training frameworks

Weak

Reskills for AI/digital era, raises TFP

+0.3–0.5%

Subtotal GDP Delta (long-run): +1.2–2.0%

6. Governance & Crisis Management Capacity

Goal: Improve institutional responsiveness and reduce political fragmentation that undermines credibility.

Component

Status

Economic Rationale

Est. GDP Impact

Simplified decision-making (e.g., majority voting for fiscal tools)

Not done

Faster response to shocks; lowers uncertainty

+0.2–0.4%

Better coordination between ECB and fiscal authorities

Fragmented

Ends reliance on monetary policy alone

+0.2–0.3%

Subtotal GDP Delta (long-run): +0.4–0.7%

TOTAL POTENTIAL LONG-RUN GDP IMPACT (compounded):

Reform Area

GDP Delta (low–high estimate)

Banking/Financial Union

+1.0% – 1.6%

Fiscal Union

+1.9% – 3.7%

Labor Integration

+0.9% – 1.3%

Regulatory Convergence

+1.8% – 2.8%

Innovation/Education

+1.2% – 2.0%

Governance Reform

+0.4% – 0.7%

Total Combined Impact

+7.2% – 12.1% of Eurozone GDP

(Note: these are long-run level effects, not annual growth rate changes.)

Final Thought: Strategic Priority Pyramid

If this were a McKinsey-style priority framework, reforms would be staged by:

Foundational: Banking union, fiscal capacity Structural: Labor mobility, regulatory convergence Productivity accelerators: R&D, digital infrastructure Enablers: Governance, political integration

So what are the respective contributions to annual growth rates for each factor?

Here’s the table that includes the currency union effect (euro adoption) alongside the other reform categories, and converts the long-run level effects into annual GDP growth rate uplifts, assuming a 10-year realization period for each component.

Eurozone Reform and Integration Strategy: Annual Growth Uplift Table (10-Year Horizon)

Reform Area

Long-Run GDP Level Gain

10-Year Annual Growth Uplift (CAGR)

Currency Union (Euro adoption)

+1.0% – 2.0%

+0.10% – 0.20%

Banking/Financial Union

+1.0% – 1.6%

+0.10% – 0.16%

Fiscal Union

+1.9% – 3.7%

+0.19% – 0.36%

Labor Mobility & Integration

+0.9% – 1.3%

+0.09% – 0.13%

Regulatory & Market Convergence

+1.8% – 2.8%

+0.18% – 0.25%

Innovation & Education Investment

+1.2% – 2.0%

+0.12% – 0.20%

Governance & Crisis Capacity

+0.4% – 0.7%

+0.04% – 0.07%

TOTAL Combined Impact

+8.2% – 14.1%

+0.79% – 1.33% per year

Baseline Eurozone Growth (1999–2019): ~1.3%/year

Post-Reform Growth Potential: ~2.1–2.6%/year, sustained for 10 years

Key Notes:

Currency union alone added modest gains (~0.1–0.2%/year) through trade, price stability, and lower interest rates. Fiscal union and regulatory convergence are the highest-yield strategies individually. All areas are complementary, not additive linearly—some interaction effects (e.g., labor mobility + banking union) reinforce gains.

Would you like a bar graph or waterfall chart visualizing this breakdown for presentation purposes?

Now putting it all together…

Unlocking the Economic Potential of the Eurozone: A Quantitative Strategy Report

Prepared by: [Your Name] | May 2025

Executive Summary

The euro, introduced in 1999, was designed to integrate European economies, reduce transaction costs, enhance price stability, and improve capital allocation. While it succeeded in political integration and macroeconomic convergence, its impact on long-term GDP growth has been modest and uneven. Empirical research finds that while core Eurozone countries gained marginally from the euro, southern economies often stagnated due to loss of competitiveness and policy flexibility.

This report assesses the actual impact of the euro, identifies the incomplete architecture of the Eurozone, and quantifies the GDP growth potential of key reforms required to fully realize the benefits of a common economic zone. The analysis includes annual growth rate boosts assuming a 10-year realization period, McKinsey-style.

1. Economic Impact of the Euro vs. Counterfactuals

Empirical Findings

Study

Method

Countries

Period

Euro Effect on GDP

Peterson Institute (2013)

Synthetic Control

Core EZ

1999–2012

~0.5–1.0 pp lower vs. counterfactual

Campos et al. (2019)

Synthetic Control

11 Euro adopters

1993–2013

+5–10% cumulative

IMF (2017)

Panel

EA19

1999–2016

Small positive for core, negative divergence for periphery

Eichengreen & Boltho (2008)

Narrative

Core EZ

1999–2007

Negligible

Channel Analysis

Channel

Positive Effect

Negative Effect

Trade Integration

+5–15% intra-EA trade

Limited GDP impact

Interest Rate Convergence

Lower borrowing costs

Credit booms, housing bubbles

Price Stability

Reduced inflation volatility

One-size-fits-all ECB policy

Capital Allocation

Better for core

Capital flight risk in crises

GDP Growth Uplift from Euro Alone

Long-run GDP level gain: +1.0% to +2.0% Annual growth uplift over 10 years: +0.10% to +0.20%

2. Beyond the Euro: Key Reforms Needed

To fully realize the economic potential of the Eurozone, reforms must go far beyond monetary integration. Below is a breakdown of reform areas, their long-run GDP impact, and annual growth rate uplift, assuming a 10-year implementation arc:

Eurozone Reform and Integration Strategy: Annual Growth Uplift Table

Reform Area

Long-Run GDP Level Gain

10-Year Annual Growth Uplift (CAGR)

Currency Union (Euro adoption)

+1.0% – 2.0%

+0.10% – 0.20%

Banking/Financial Union

+1.0% – 1.6%

+0.10% – 0.16%

Fiscal Union

+1.9% – 3.7%

+0.19% – 0.36%

Labor Mobility & Integration

+0.9% – 1.3%

+0.09% – 0.13%

Regulatory & Market Convergence

+1.8% – 2.8%

+0.18% – 0.25%

Innovation & Education Investment

+1.2% – 2.0%

+0.12% – 0.20%

Governance & Crisis Capacity

+0.4% – 0.7%

+0.04% – 0.07%

TOTAL Combined Impact

+8.2% – 14.1%

+0.79% – 1.33% per year

3. Component-by-Component Strategy

A. Banking & Capital Markets Union

SSM and SRM in place, but EDIS remains incomplete. Full integration would equalize borrowing costs, reduce financial fragmentation. GDP Impact: +1.0% – 1.6% | Annual Growth Uplift: +0.10% – 0.16%

B. Fiscal Union

EU budget is ~1% of GDP vs. U.S. federal ~20%. Needs: permanent fiscal stabilization tools, common bonds, automatic transfers. GDP Impact: +1.9% – 3.7% | Annual Growth Uplift: +0.19% – 0.36%

C. Labor Market Integration

Legal, pension, housing, and language barriers remain high. Poor mobility limits rebalancing across regions. GDP Impact: +0.9% – 1.3% | Annual Growth Uplift: +0.09% – 0.13%

D. Regulatory & Market Convergence

Services directive and digital market remain under-implemented. Huge untapped gains in productivity and scale effects. GDP Impact: +1.8% – 2.8% | Annual Growth Uplift: +0.18% – 0.25%

E. Innovation & Education Investment

Horizon Europe helps, but spending still lags China and U.S. Broadband, AI, 5G rollouts and workforce retraining are key. GDP Impact: +1.2% – 2.0% | Annual Growth Uplift: +0.12% – 0.20%

F. Governance Reforms

Crisis response hindered by unanimity voting and legal complexity. Stronger central institutions reduce uncertainty and risk premiums. GDP Impact: +0.4% – 0.7% | Annual Growth Uplift: +0.04% – 0.07%

4. Strategic Outlook: Growth Potential and Prioritization

If all reforms are executed within a 10-year time frame, the Eurozone’s average real GDP growth could rise from ~1.3% to ~2.1–2.6% per year, bringing it closer to U.S. and high-performing East Asian standards.

Reform Priority Pyramid

Tier

Reforms

Foundational

Fiscal union, banking union, governance mechanisms

Structural

Regulatory convergence, labor mobility

Productivity Accelerators

R&D, digital investment, education

Enablers

Political harmonization, public support infrastructure

Conclusion

The euro laid the groundwork for integration, but it remains an incomplete project. Structural reforms across fiscal policy, finance, labor, and innovation could unlock a compounded +8–14% gain in GDP, translating into +0.8 to +1.3 percentage points of additional annual growth over a decade.

This would transform the Eurozone from a low-growth, institutionally fragmented bloc into a globally competitive and economically resilient union.

And even a PPT

https://docs.google.com/presentation/d/18ONXIBAd_lZagI6ADwr7ud4Kbu6e4vdp/edit?usp=drivesdk&ouid=109714124272825709671&rtpof=true&sd=true

And impressive how well ChatGPT did it all, no?

Using a top-down, Fermi-style estimation approach, here’s the result:

Total traffic hours saved worldwide (2010–2020): approximately 185.6 billion hours Total estimated economic value gained from these time savings: approximately $1.86 trillion USD

This assumes:

A global vehicle fleet of 1.5 billion vehicles, 300 driving hours per vehicle per year, Navigation app adoption rising from 10% to 90% over 10 years, An average of 7.5% time savings for users, And a global average value of time of $10/hour.

Scroll down for the stuff that sucks by comparison. But first the assumptions.

1. Global Number of Vehicles: ~1.5 Billion

Source: International Organization of Motor Vehicle Manufacturers (OICA) and World Bank reports In 2020, the global number of registered vehicles was estimated at around 1.4 to 1.5 billion. Growth is driven primarily by developing nations, with saturation in high-income countries.

Reference:

OICA Statistics

2. Average Hours Driven per Vehicle per Year: ~300 Hours

Basis: In the U.S., average annual miles per vehicle is about 12,000 miles, at an average speed of ~35 mph → ~340 hours/year. In Europe and many OECD countries, the number is lower, ~10,000 km/year. In developing countries, usage is generally lower (fewer commutes, fuel costs, etc.).

So a global weighted average of ~300 hours/year per vehicle is a reasonable conservative estimate.

3. App Adoption: 10% (2010) to 90% (2020)

Basis: Google Maps launched turn-by-turn navigation in 2009–2010. Waze, acquired by Google in 2013, accelerated user adoption of crowdsourced routing. By the late 2010s, smartphones had 70–90% penetration in developed markets. In developing countries, adoption lagged but grew significantly due to Android.

References:

Statista: Smartphone adoption worldwide Pew Research: Mobile Technology Use

4. Time Savings from Apps: ~7.5%

Basis: Field studies suggest 5–10% reductions in average travel time from GPS-based routing. Smart routing helps users avoid peak congestion, accidents, and construction delays. A study in Pittsburgh found adaptive signal control reduced travel time by 25%; routing apps likely deliver a more modest but widespread improvement.

References:

ASCE 2021: Impact of Real-Time Traffic Apps Pittsburgh SURTRAC Project

5. Value of Time: $10/hour Global Average

Basis: OECD and World Bank estimates of “value of statistical life” (VSL) and time use often cite: ~$15–$30/hr in high-income countries ~$2–$5/hr in lower-income countries A global weighted average of ~$10/hour is defensible for private time savings.

References:

World Bank: The Value of Time in Economic Evaluation OECD: Cost-Benefit Analysis and the Value of Time

Stuff that sucks—

Many popular consumer inventions offer high subjective enjoyment or convenience but relatively low economic productivity gains. Here are a few inventions or products people love that generally create less economic value per hour used than things like traffic apps:

1. Social Media Platforms (e.g., Instagram, TikTok, Facebook)

Enjoyment: Very high — billions of users. Economic Value: Low per hour — much time spent passively consuming content. Downside: Opportunity cost of attention/time, mental health concerns.

2. Streaming Services (e.g., Netflix, Disney+, YouTube for leisure)

Enjoyment: High entertainment value. Economic Value: Mostly leisure consumption — little productivity impact. Value-to-Cost Ratio: Excellent for consumers, but limited macroeconomic impact.

3. Video Games (especially mobile or casual games)

Usage: Hundreds of billions of hours spent globally. Economic Impact: Generates industry revenue but limited cross-sector productivity. Note: Competitive gaming and game development do add economic value — but far less per user-hour than infrastructure or navigation tech.

4. Voice Assistants (e.g., Alexa, Siri)

Convenience: Popular for music, timers, and light control. Economic Gain: Marginal productivity boost — few significant time savings versus manual use.

5. Smart Home Devices (e.g., smart lights, fridges, vacuums)

Perceived Innovation: High. Actual Impact: Often marginal time savings or novelty with low return on cost.

6. VR Headsets (e.g., Oculus/Meta Quest)

Innovation: Technically impressive and loved by users. Economic ROI: Very low so far — adoption and productivity use cases remain niche.

7. E-Scooters / Micro-Mobility Apps

Love Factor: Fun, trendy in urban areas. Net Impact: Questionable — can substitute walking or transit, and often require city subsidies or cleanup.

Disco kind of blew up, got submerged, and then all dance got reinvented underground in warehouses (“house”), merged with electronic sounds (Kraftwerk) and later pop (Madonna?), rock (Happy Mondays? Postal Service?) and hip hop and at this point it’s basically all music.

Here’s a great history-of doc:

Here’s an insanely great discussion of Rockism, early R&B vs. Rock vs. Soul, and more from the New Yorker writer Kelefah Saleh: podcast

And a classic and influential music criticism piece he wrote about Rockism vs. Poptimism: NYTimes

And here’s a summary of all that above including a playlist at the bottom:

Detailed History of House Music

House music emerged in the late 1970s and early 1980s from the underground clubs of Chicago, evolving from the ashes of disco and becoming a global phenomenon. Below is a summarized history of house music as described in the provided transcript, focusing on its origins, evolution, key milestones, and cultural impact.

Origins in Chicago (Late 1970s–Early 1980s)

• Roots in Disco and Underground Clubs: House music was born in Chicago, heavily influenced by disco, which had been kept alive by gay and Black communities in underground clubs after its mainstream decline following the 1979 anti-disco backlash led by DJ Steve Dahl. Disco’s danceable R&B rhythms formed the foundation for house.
• The Warehouse and Frankie Knuckles: In 1977, club promoter Robert Williams brought Frankie Knuckles, a New York DJ, to Chicago to play at his new club, the Warehouse. Knuckles introduced New York’s gay club energy, playing soulful, uplifting disco classics and underground tracks like Loose Joints’ “Is It All Over My Face” and Anita Ward’s “Ring My Bell.” The term “house music” originated from the music played at the Warehouse, initially referring to old disco and soul records, later shortened to “house” by local record shops like Imports Etc.
• Paradise Garage and Larry Levan Influence: In New York, the Paradise Garage, opened in 1978, was a pivotal influence with its resident DJ Larry Levan. Levan’s eclectic sets, blending disco, punk, and various rhythms, created a legendary club atmosphere that inspired Chicago’s house scene. His ability to control crowds with music and set trends influenced house DJs.

Evolution of House Music (1980s)

• Chicago’s DJ Innovations: Chicago DJs, facing a limited supply of disco records post-backlash, became highly creative. They developed tight mixing techniques, repeating and manipulating record sections to keep crowds engaged. The Hot Mix Five on WBMX radio popularized this style, playing short, impactful segments of records and using razor blades to edit reel-to-reel tapes for creative mixes.
• Drum Machines and Original Tracks: By the early 1980s, DJs began using drum machines to create their own beats, giving house a harder edge. This marked the shift from recycling disco to producing original music. Competition among DJs like Steve “Silk” Hurley and Farley “Jackmaster” Funk intensified, with each aiming to create the hottest tracks.
• The Music Box and Ron Hardy: In 1983, the Music Box club opened in Chicago, with DJ Ron Hardy as its star. Hardy’s intense, loud sets, often sped up due to his heroin use, took clubbing to new heights. His club was a cultural hub where “jacking”—a provocative dance style—became synonymous with house culture.
• First House Records: In 1984, Jamie Principle’s demo tape “Your Love” spread virally through Chicago clubs, marking a turning point with its hypnotic, original sound. In 1983, Jesse Saunders and Vince Lawrence released “On and On,” the first house record on vinyl, sparking the Chicago house boom by showing non-musicians they could produce music. Larry Sherman’s Trax Records, launched in the mid-1980s, capitalized on this wave, pressing records quickly using recycled vinyl, though quality was often poor.

Acid House and Global Spread (Mid-1980s)

• Birth of Acid House: In late 1985, DJ Pierre and friends created acid house by experimenting with the Roland TB-303, producing the track “Acid Tracks.” Ron Hardy played it multiple times at the Music Box, cementing its popularity. Pierre’s musical approach to the 303 distinguished it from later imitations.
• Marshall Jefferson’s Contributions: Marshall Jefferson added sophistication to house with tracks like “Move Your Body” (1986), dubbed the “house music anthem.” Its piano-driven sound initially faced resistance but became internationally recognized, giving house a name and identity.
• UK Adoption and Summer of Love (1986–1988): House crossed the Atlantic in 1986, with tracks like Farley “Jackmaster” Funk’s “Love Can’t Turn Around” reaching the UK Top 10. Northern England’s Northern Soul fans embraced house’s fast, soulful rhythms. The Hacienda club in Manchester, opened in 1982 by Factory Records, became a house music hub, with DJs like Mike Pickering and Graeme Park playing Chicago tracks. In 1987, British DJs Paul Oakenfold, Danny Rampling, Nicky Holloway, and Johnny Walker experienced Ibiza’s Amnesia club, where DJ Alfredo played a balearic mix of house, pop, and disco. Inspired, Rampling opened Shoom in London, recreating Ibiza’s vibe and sparking the UK’s acid house craze.
• Summer of Love (1988): The combination of house music and ecstasy transformed British club culture. Clubs like the Hacienda and Shoom, along with illegal raves around the M25, fueled the 1988 Summer of Love. Tracks like Gerald Simpson’s “Voodoo Ray” and Orbital’s “Chimes” defined the UK scene, blending house with techno and breakbeats.

Diversification and Mainstream Success (Late 1980s–1990s)

• Techno and Detroit’s Influence: In Detroit, house evolved into techno, with tracks like Inner City’s “Big Fun” (1988) and Juan Atkins’ work blending Kraftwerk and funk influences. Techno’s intellectual approach contrasted with house’s emotional core.
• Indie Dance and Rave Culture: In the UK, bands like Happy Mondays fused house rhythms with indie rock, producing tracks like “Wrote for Luck” (1988). Shut Up and Dance added breakbeats to house, laying the groundwork for jungle and UK garage. The rave scene grew from underground parties to mainstream licensed events by 1992, despite police crackdowns and the 1994 Criminal Justice Bill targeting repetitive beats.
• New Styles in the UK: By the late 1980s, ambient house (e.g., The Orb’s “Little Fluffy Clouds”) emerged, blending down-tempo beats with New Age sounds. Hardcore, jungle, and drum and bass evolved from faster tempos and sophisticated breakbeats, with Goldie’s “Timeless” (1995) bringing drum and bass to the charts. Leftfield’s fusion of house, techno, and dub, especially with John Lydon’s “Open Up” (1993), crossed into mainstream pop.
• US Mainstream Breakthrough: In the US, house remained underground until the 1990s, when producers like Masters at Work and Armand Van Helden remixed pop tracks (e.g., Tori Amos’ “Professional Widow”) into club hits, elevating DJs to studio producers. Todd Terry’s hip-hop-influenced house, with quick samples and rugged basslines, set a blueprint for modern house.

Continued Evolution (2000s and Beyond)

• UK Garage: By 2001, UK garage emerged as a distinctly British sound, rooted in house but incorporating reggae basslines and sped-up US tracks by producers like Todd Edwards. It reflected urban UK youth culture, with kids making music in bedrooms.
• Global Phenomenon: House music’s ability to reinvent itself through sampling and innovation made it a global force, crossing racial, sexual, and class divides. Its emotional core—described as a “feeling” that moves dance floors—ensured its longevity.

Key Figures Mentioned in the Transcript

• Frankie Knuckles: The “Godfather of House,” brought New York’s gay club energy to Chicago’s Warehouse, defining the early house sound with soulful disco classics.
• Larry Levan: Paradise Garage’s legendary DJ, whose eclectic sets and crowd control inspired house’s emotional intensity.
• Ron Hardy: Music Box DJ known for intense, loud sets that popularized acid house and jacking culture.
• Jamie Principle: Created “Your Love” (1984), a seminal house track that marked the genre’s shift to original compositions.
• Jesse Saunders and Vince Lawrence: Produced “On and On” (1983), the first house record on vinyl, sparking the Chicago house boom.
• DJ Pierre: Pioneered acid house with “Acid Tracks” (1985), using the Roland TB-303 in a musical way.
• Marshall Jefferson: Added sophistication with “Move Your Body” (1986), the house music anthem, and formed Ten City, a house supergroup.
• Farley “Jackmaster” Funk and Steve “Silk” Hurley: Competitive DJs who produced early hits like “Love Can’t Turn Around” (1986).
• Larry Sherman: Trax Records owner, pressed early house records quickly, though often with poor quality due to recycled vinyl.
• Adonis: Contributed to the emerging house sound with tracks like “No Way Back” (1986).
• Larry Heard: Chicago producer, part of the 1986 UK tour, known for deep house contributions.
• Joe Smooth: Created “Promised Land” (1988), a house anthem inspired by the UK tour.
• Todd Terry: New York producer who brought hip-hop sensibilities to house, influencing modern production.
• Paul Oakenfold, Danny Rampling, Nicky Holloway, Johnny Walker: British DJs who brought Ibiza’s balearic house to London, sparking the Summer of Love.
• DJ Alfredo: Ibiza’s Amnesia DJ, whose balearic mix influenced UK house.
• Mike Pickering and Graeme Park: Hacienda DJs who popularized Chicago house in Manchester.
• Gerald Simpson: Produced “Voodoo Ray” (1988), a UK house classic.
• Orbital: Created “Chimes” (1989), blending house and techno, mastering UK dance music.
• Happy Mondays: Fused indie rock with house in “Wrote for Luck” (1988), expanding house’s audience.
• Shut Up and Dance: Added breakbeats to house, influencing jungle and UK garage.
• Goldie: Pioneered drum and bass with “Timeless” (1995), evolving house’s rhythms.
• Leftfield: Blended house, techno, and dub, collaborating with John Lydon on “Open Up” (1993).
• Armand Van Helden: Remixed pop tracks like Tori Amos’ “Professional Widow” (1996), bringing house to US charts.
• Masters at Work: Remixed pop into club hits, elevating DJs as producers.
• Todd Edwards: Influenced UK garage with his chopped vocal samples.

Playlist of Songs Featured in the Transcript

The following table lists all songs mentioned in the transcript, their artists, the era they represent, and a brief characterization of their contribution to house music’s evolution. Amol’s Apple playlist.

1. Robert Longo

“If Longo hadn’t come along, his art could have been dreamed up by a team of artists at a Madison Avenue advertising agency.”

2. Cecily Brown

“Pointlessly messy, made to sell, and uninteresting from any distance.”

3. Banks Violette

“Another example of macho posturing masquerading as art, cloaked in gothic doom and empty theatricality.”

4. Damien Hirst

“Swagger and sideburns art… All flash and no lasting resonance.”

5. Gabriel Orozco

“Dispiritingly one-note… visually austere to the point of nullity.”

6. Roni Horn

“A one-note minimalist sensibility that feels cold and emotionally paralyzed.”

7. Urs Fischer

“Another exercise in clever emptiness, suffocated by its own whimsy.”

8. Tino Sehgal

“Emotionally inert… performances that vanish before they register.”

9. Mickalene Thomas

“Fast-food obviousness — you get it too quickly and then it’s over.”

10. Josephine Meckseper

“How much did this thing cost? All gloss, no gravity.”

So was she predictive?

1. Robert Longo

Featured in The Pictures Generation exhibition at the Metropolitan Museum of Art (2009)

2. Cecily Brown

Major retrospective at the Metropolitan Museum of Art: Cecily Brown: Death and the Maid (2023)

3. Banks Violette

Solo exhibition at the Whitney Museum of American Art (2005)

4. Damien Hirst

Retrospective at Tate Modern: Damien Hirst (2012)

5. Gabriel Orozco

Retrospective at the Museum of Modern Art, New York (2009)

6. Roni Horn

Major solo exhibition at Tate Modern: Roni Horn aka Roni Horn (2009)

7. Urs Fischer

Solo exhibition at the Guggenheim Museum: Urs Fischer: Marguerite de Ponty (2009)

8. Tino Sehgal

Winner of the Golden Lion at the Venice Biennale (2013)

9. Mickalene Thomas

Retrospective at the Brooklyn Museum: Origin of the Universe (2012)

10. Josephine Meckseper

Platform commission and solo exhibition at the Parrish Art Museum (2013)

From one of my pilgrimages to Omaha.

Every other nonsense idea seems to be thriving at the moment. Shouldn’t RFK Jr be campaigning against it?

Mobile phones and autism rates have grown at the same rate since 1998.

Key Learnings and Insights from the LifeX AI x Bio Conference

Foundation Models in Biology

• There is some debate on the exact definition of a foundation model.
• A key characteristic is formulating tasks in an open-ended way, throwing large amounts of unstructured data at a model to see what it can do.
• Foundation models are good at many tasks, even those they weren’t specifically designed to do.
• The term “foundation” suggests serving as a base for multiple applications; it may also signal importance for fundraising.
• Thomas at Vanta AI works on all-atom foundation models for protein-small molecule complexes. These models are expensive to train, but not necessarily $500M expensive.
• Compared to LLMs trained on exabytes of text, biological data models (e.g., AlphaFold) use far less—terabytes at most.
• Biodata could be nearly infinite if gathered and organized; generating and curating data is essential.
• Multimodal models are emerging, but their advantages remain debated.
• Surprising Insight: Sean suggests AlphaFold may be the only “true” foundation model in biomedicine, showing emergent properties.
• LLMs can correlate vision and text to describe unseen images—this is analogous to AlphaFold’s capabilities.
• Multimodal models exist for tasks like disease diagnosis, but true discipline-spanning foundation models remain rare.
• Agent-based systems may one day abstract away model selection entirely.
• Medical specialties may be social/reimbursement boundaries, not scientific ones—unlike the medicine–physics divide.

Data in Biology and AI

• Data is more important than model architecture in biomedicine today.
• Data hoarding is common due to its value.
• Compared to finance, bio shows more anxiety around data sharing.
• Surprising Insight: HIPAA concerns drive biopharma to run models on-prem, even with third-party data—distinct from regulation, it’s about ownership fear.
• Data quality issues persist, especially in legacy phenotypic datasets.
• Labeled data is a frontier—time-consuming to create but crucial.
• Outlier biological traits (e.g., insensitivity to pain) are valuable for discovery.
• Synthetic data can help; sometimes it outperforms radiologists, but quality is bounded by generation process.
• Data-scarce strategies include simulation (e.g., virtual cell modeling).
• Mechanistic models are deterministic but don’t reduce compute burden.
• Depth > breadth: Germline mutation data can offer stronger predictive power than more shallow data.

Infrastructure

• Options: pay for APIs, self-host in cloud, or buy GPUs.
• AWS hosts many models; Bedrock aims to simplify access for bio use cases.
• Goal: Make complex platforms usable for less technical scientists.
• Bio infrastructure mixes cloud (scalable) and on-prem (privacy).
• Training models is costly; some companies cloud-hop for credits. Inference is increasingly expensive too.

AI for Accelerating Drug Discovery

• The panel discussed reversing Irum’s Law (rising drug discovery costs).
• Challenges include combinatorial therapy and multi-target inhibition.
• AI is used for hypothesis generation and predicting experimental outcomes.
• AI may replace some experimental work, boosting iteration speed.
• Surprising Insight: Only ~10% of drugs entering the clinic succeed—improving this is AI’s big opportunity.
• The future lies in causal biology—differentiated hypotheses via mechanistic insights.
• AI also helps in clinical ops: patient selection, endpoints, data management.
• New architectures include flow matching, agentic systems, RL, and chain-of-thought reasoning.
• Fragmented data across stages limits end-to-end modeling; bridging these gaps could drive the next wave.
• Pharma’s in vivo/in vitro data is valuable for improving translation.
• Surprising Insight: Biotech “sells hope” to pharma, pharma to patients, investors buy hope—this drives business models.
• Some startups sell predictions (anti-hope) vs. molecules (hope).
• Hype is seen as necessary to attract capital and talent.
• Should vertical AI tools build pipelines (“platform envy”) or stay focused and share in upside?

Longevity – Hype vs. Reality

• Longevity focuses on aging drivers, not just disease prevention.
• Hallmarks include mitochondrial dysfunction and chronic inflammation.
• Factions focus on specific aging mechanisms (e.g., mitochondria, immune).
• Measuring dysfunction is tough outside research settings.
• Defining “health span” shapes longevity goals.
• The space has evolved from cosmetic/optimization to biochemistry-based interventions.
• Surprising Insight/Debunking: Brian Johnson’s N=1 experiments are viewed skeptically; epigenetic clocks have high variability.
• Interventions like NAD, metformin, rapamycin, fasting are unproven for lifespan extension.
• Uncontrolled self-experimentation may have unknown harms.
• Antioxidants could worsen some cancers in mice.
• Surprising Insight: Plasma exchange therapy shows promise in mice but isn’t practical/safe for humans.
• Validated biomarkers remain the standard; most physicals miss deeper aging indicators.
• Biomarkers help both stratify patients and track intervention impact.
• Best validated interventions: eat less, sleep more, be happy, exercise.
• Fasting is a consistent positive signal.
• GLP-1s could be a “miracle drug” for longevity, pending long-term studies.
• Mother–daughter asymmetry shows rejuvenation is biologically possible and worth studying.

Starting a Company as a PhD

• Surprising Insight: Great academics often make weak founders; bold simplification is punished in academia but necessary in startups.
• Startups require prioritization, speed, and clarity.
• Choose the right seed investor, not the first.
• Bootstrapping can prove pull before raising.
• Serendipitous conversations lead to funding opportunities.
• Startup law firms can defer fees and provide crucial early help.
• Trust and team empowerment are critical.
• University IP can be a landmine—build off-site, license when needed.
• Cofounder questionnaires help surface conflicts early.
• Biotech companies tend to delay hiring GCs until IPO stage or regulatory complexity arises.
• Academic rigor (p-values, caveats) must give way to rapid, imperfect action.
• Simplifying your pitch is essential.
• Conviction comes from personal connection to the problem and end user.
• Surprising Insight: Standard startup structures (stock, boards, investor terms) exist for a reason—over-customization creates drag.
• Legal simplicity allows focus on harder problems.

Key Learnings and Insights from the LifeX AI x Bio Conference

Welcome and Overview

• The event was one of several AI and Bio conferences co-hosted by Fenwick & West and LifeX Ventures.
• Fenwick & West is a tech and life sciences focused law firm supporting venture-backed companies and investors.
• LifeX Ventures is an early-stage venture fund (Seed and Series A) focused on “AI Meets Bio,” specifically pure tech companies where the customer is a scientist or doctor, aiming to make things better so people live longer.
• LifeX Ventures has made 55–56 deals in about two and a half years.

Foundation Models in Biology

• There is some debate on the exact definition of a foundation model.
• A key characteristic is formulating tasks in an open-ended way, throwing large amounts of unstructured data at a model to see what it can do.
• Foundation models are good at many tasks, even those they weren’t specifically designed to do.
• The term “foundation” suggests serving as a base for multiple applications; it may also signal importance for fundraising.
• Thomas at Vanta AI works on all-atom foundation models for protein-small molecule complexes. These models are expensive to train, but not necessarily $500M expensive.
• Compared to LLMs trained on exabytes of text, biological data models (e.g., AlphaFold) use far less—terabytes at most.
• Biodata could be nearly infinite if gathered and organized; generating and curating data is essential.
• Multimodal models are emerging, but their advantages remain debated.
• Surprising Insight: Sean suggests AlphaFold may be the only “true” foundation model in biomedicine, showing emergent properties.
• LLMs can correlate vision and text to describe unseen images—this is analogous to AlphaFold’s capabilities.
• Multimodal models exist for tasks like disease diagnosis, but true discipline-spanning foundation models remain rare.
• Agent-based systems may one day abstract away model selection entirely.
• Medical specialties may be social/reimbursement boundaries, not scientific ones—unlike the medicine–physics divide.

Data in Biology and AI

• Data is more important than model architecture in biomedicine today.
• Data hoarding is common due to its value.
• Compared to finance, bio shows more anxiety around data sharing.
• Surprising Insight: HIPAA concerns drive biopharma to run models on-prem, even with third-party data—distinct from regulation, it’s about ownership fear.
• Data quality issues persist, especially in legacy phenotypic datasets.
• Labeled data is a frontier—time-consuming to create but crucial.
• Outlier biological traits (e.g., insensitivity to pain) are valuable for discovery.
• Synthetic data can help; sometimes it outperforms radiologists, but quality is bounded by generation process.
• Data-scarce strategies include simulation (e.g., virtual cell modeling).
• Mechanistic models are deterministic but don’t reduce compute burden.
• Depth > breadth: Germline mutation data can offer stronger predictive power than more shallow data.

Infrastructure

• Options: pay for APIs, self-host in cloud, or buy GPUs.
• AWS hosts many models; Bedrock aims to simplify access for bio use cases.
• Goal: Make complex platforms usable for less technical scientists.
• Bio infrastructure mixes cloud (scalable) and on-prem (privacy).
• Training models is costly; some companies cloud-hop for credits. Inference is increasingly expensive too.

AI for Accelerating Drug Discovery

• The panel discussed reversing Irum’s Law (rising drug discovery costs).
• Challenges include combinatorial therapy and multi-target inhibition.
• AI is used for hypothesis generation and predicting experimental outcomes.
• AI may replace some experimental work, boosting iteration speed.
• Surprising Insight: Only ~10% of drugs entering the clinic succeed—improving this is AI’s big opportunity.
• The future lies in causal biology—differentiated hypotheses via mechanistic insights.
• AI also helps in clinical ops: patient selection, endpoints, data management.
• New architectures include flow matching, agentic systems, RL, and chain-of-thought reasoning.
• Fragmented data across stages limits end-to-end modeling; bridging these gaps could drive the next wave.
• Pharma’s in vivo/in vitro data is valuable for improving translation.
• Surprising Insight: Biotech “sells hope” to pharma, pharma to patients, investors buy hope—this drives business models.
• Some startups sell predictions (anti-hope) vs. molecules (hope).
• Hype is seen as necessary to attract capital and talent.
• Should vertical AI tools build pipelines (“platform envy”) or stay focused and share in upside?

Longevity – Hype vs. Reality

• Longevity focuses on aging drivers, not just disease prevention.
• Hallmarks include mitochondrial dysfunction and chronic inflammation.
• Factions focus on specific aging mechanisms (e.g., mitochondria, immune).
• Measuring dysfunction is tough outside research settings.
• Defining “health span” shapes longevity goals.
• The space has evolved from cosmetic/optimization to biochemistry-based interventions.
• Surprising Insight/Debunking: Brian Johnson’s N=1 experiments are viewed skeptically; epigenetic clocks have high variability.
• Interventions like NAD, metformin, rapamycin, fasting are unproven for lifespan extension.
• Uncontrolled self-experimentation may have unknown harms.
• Antioxidants could worsen some cancers in mice.
• Surprising Insight: Plasma exchange therapy shows promise in mice but isn’t practical/safe for humans.
• Validated biomarkers remain the standard; most physicals miss deeper aging indicators.
• Biomarkers help both stratify patients and track intervention impact.
• Best validated interventions: eat less, sleep more, be happy, exercise.
• Fasting is a consistent positive signal.
• GLP-1s could be a “miracle drug” for longevity, pending long-term studies.
• Mother–daughter asymmetry shows rejuvenation is biologically possible and worth studying.

Starting a Company as a PhD

• Surprising Insight: Great academics often make weak founders; bold simplification is punished in academia but necessary in startups.
• Startups require prioritization, speed, and clarity.
• Choose the right seed investor, not the first.
• Bootstrapping can prove pull before raising.
• Serendipitous conversations lead to funding opportunities.
• Startup law firms can defer fees and provide crucial early help.
• Trust and team empowerment are critical.
• University IP can be a landmine—build off-site, license when needed.
• Cofounder questionnaires help surface conflicts early.
• Biotech companies tend to delay hiring GCs until IPO stage or regulatory complexity arises.
• Academic rigor (p-values, caveats) must give way to rapid, imperfect action.
• Simplifying your pitch is essential.
• Conviction comes from personal connection to the problem and end user.
• Surprising Insight: Standard startup structures (stock, boards, investor terms) exist for a reason—over-customization creates drag.
• Legal simplicity allows focus on harder problems.

I paid $150

Their cost for NY->CDMX…?

Notably, many workplace coffee machines use metal filters or lack proper filtration, leading to higher diterpene content in the coffee they produce. This suggests that office coffee may contribute more to elevated cholesterol levels than home-brewed, paper-filtered coffee. 

However, some studies indicate that even paper-filtered coffee can modestly increase LDL cholesterol and inflammation markers, regardless of roast level. 

In summary, while paper-filtered coffee is generally preferable for managing cholesterol, moderation remains important, especially for individuals with cardiovascular risk factors.

Deceleration Without Collapse: Trade, Power, and Possibility in a Post-Growth World

In the wake of planetary ecological crises, political paralysis, and widespread disillusionment with both neoliberalism and authoritarian statism, a new strand of economic and political thought has emerged: decelerationism, often aligned with the broader degrowth movement. This intellectual current rejects the fantasy that economic growth—whether driven by green technologies or digital efficiencies—can be decoupled from ecological destruction. Instead, it calls for a managed slowing down of industrial throughput, energy use, and material production, combined with new social relations grounded in sufficiency, autonomy, and care.

A central figure in this movement is Kohei Saito, whose books Capital in the Anthropocene and Slow Down have captured wide popular attention in Japan and beyond. Saito blends ecological Marxism, anti-productivism, and a bottom-up communitarian political philosophy to articulate a vision of post-capitalist prosperity without growth. Unlike Green New Dealers who emphasize state-led investment in renewables and infrastructure, or accelerationists who seek emancipation through runaway technological transformation, Saito advocates for a deliberate, democratic contraction of the economy—a slowdown from below.

But if growth is out, and centralized state planning is suspect, what kind of coordination, exchange, and governance might still be possible at planetary scale?

I. Saito’s Critique: Growth, Green Deals, and Techno-Futures

Saito’s main targets include:

• Green Keynesianism / Green New Deal policies: These rely on the state to stimulate “green growth” through massive investment in renewable energy, infrastructure, and jobs. Saito argues this merely recycles productivist assumptions, shifting the carbon problem to the mining of lithium, rare earth metals, and global supply chains.
• Techno-accelerationism: Common among Silicon Valley futurists and left-wing Prometheans, this view sees automation, AI, and geoengineering as the solutions to planetary constraints. Saito sees this as a delusion of control—one that accelerates inequality and ecological overshoot.
• Mainstream Marxism and socialism: He criticizes traditional leftist belief in industrial growth and state planning, calling for a rupture from productivist Marxist legacies, especially those rooted in 20th-century industrial socialism.

II. Degrowth, Deceleration, and Political Power

Saito’s approach contrasts with both state-centric and market-centric models. Instead, he promotes:

• Commons-based production: Local control of land, tools, and labor, governed democratically and aimed at meeting basic needs, not maximizing output.
• Community autonomy: Power devolved to self-organized, interdependent groups—drawing inspiration from cooperatives, eco-villages, and indigenous modes of governance.
• Use-value over exchange-value: Economies rooted in care, maintenance, and reproduction, not profit or competition.

This represents a form of post-capitalist municipalism, akin to what Murray Bookchin, David Graeber, and Silvia Federici have described: localized control, non-hierarchical relations, and feminist and ecological ethics of sufficiency.

III. Trade in a Degrowth World: Problem or Possibility?

A persistent tension in degrowth thought is the role of trade and exchange. If communities are producing what they need locally, what happens to non-local goods—tools, spices, medicines, metals, or knowledge? Does interdependence inevitably reintroduce capitalist dynamics, or is another form of global exchange possible?

Degrowth-compatible trade principles:

• Biophysical quotas: Trade is limited not by market demand, but by planetary boundaries (e.g., carbon, land use, energy).
• Embedded emissions and labor standards: Goods traded only if their production meets strict social and ecological criteria.
• Reciprocity and mutual sufficiency: Exchange based on meeting needs, not extracting surplus.
• Non-market institutions: No reliance on prices, competition, or financial speculation—trade mediated via protocols, reputation, and social rules.

Precedents and analogues:

• Solidarity economies in Latin America (e.g., barter clubs in Argentina, mercados solidarios in Ecuador).
• Fair trade co-ops and food sovereignty networks, especially in indigenous and peasant movements.
• Digital commons and open source licensing (e.g., Creative Commons, GNU, Open Source Ecology).
• Cosmopolitan localism (glocalism): Ezio Manzini’s model where communities are autonomous but interlinked through transparent, ethical exchange.

IV. Degrowth WTO? Rethinking Global Governance

A provocative thought experiment: what would a WTO-for-degrowth look like? Instead of facilitating GDP growth, such an institution would coordinate ecological justice, redistribution, and biophysical limits.

Features might include:

• Global carbon/resource caps: Country-level trade quotas based on emissions or material throughput.
• Ecological tribunals: Disputes resolved not by market distortion claims, but environmental harm and social exploitation.
• Mutual aid treaties: Agreements for knowledge sharing, reparations, and redistribution—especially from historically high-consuming countries.
• Local priority clauses: Ensuring communities can prioritize local production without penalties for “trade distortion.”
• Commons Exchange Authority: Supporting barter, mutual credit systems, and non-capitalist forms of value transfer.

V. The Authoritarian Risk: Rulemaking Without Markets?

But such a system raises serious questions: without price signals or competitive markets, how are decisions made? Who sets quotas, enforces limits, and adjudicates conflicts?

Dangers of centralization:

• Technocracy: Environmental accounting and quota systems could become tools of elite governance, opaque to most people.
• Normative monopoly: Who defines what is “enough” or “fair”? Without pluralism, universalist frameworks can impose particular ideologies.
• Bureaucratic creep: Managing sufficiency, carbon budgets, and embedded emissions may lead to surveillance, complexity, and inertia.

Saito resists top-down planning in favor of federated autonomy. But this only deepens the coordination challenge: how can we avoid both authoritarianism and fragmentation?

VI. Analogies from Non-Economic Systems

Surprisingly, there are functioning large-scale systems outside of economics that manage complexity without central control or profit motives:

1. Open Source Software

• Voluntary, decentralized collaboration.
• Coordination via protocols, version control, and shared norms (e.g. Git, Linux).
• Maintains quality and innovation without markets.

2. Scientific Research Networks

• Peer review, open publishing, and citation economies.
• Distributed expertise, global collaboration, adaptive problem solving.
• Fast and effective during COVID-19 vaccine development.

3. Jazz Improvisation Ensembles

• Emergent coherence via mutual listening and shared rhythm.
• Leadership is temporal, fluid, and context-dependent.
• Local autonomy within bounded structures.

4. Decentralized Monastic Networks

• Self-governing communities with shared values (e.g. Benedictine federations, Buddhist sanghas).
• Resilience through nested autonomy and spiritual purpose.

5. Commons Governance (Elinor Ostrom)

• Resource management through local rules, monitoring, and sanctions.
• Polycentric governance scales without hierarchy.
• Proven in forests, fisheries, and irrigation systems.

These examples suggest that distributed, value-aligned, protocol-driven systems can coordinate at scale without coercion or markets.

VII. Reimagining Coordination: Protocols, AI, and the Moral Economy

To scale degrowth without returning to command economies or growth logics, we need new forms of coordination—technological, cultural, and moral.

Possibilities:

• Distributed ledgers and DAOs: Use blockchain not for speculation, but for ecological accounting and commons governance.
• AI as planning infrastructure: Adaptive, transparent tools for managing flows of labor, material, and energy across federated systems.
• Moral currencies: Measures of sufficiency, care, or ecological repair to replace or supplement prices.
• Anarchist and Indigenous governance: Deep traditions of non-state coordination rooted in consensus, ritual, and reciprocity.

VIII. Conclusion: A Cosmopolitan Sufficiency

Kohei Saito’s vision pushes us to rethink the deepest assumptions of modern economics: that prosperity requires growth, that coordination requires hierarchy, and that globalization requires capital. The challenge now is not merely to reject these premises—but to build functioning, liberatory alternatives.

A future of sufficiency, solidarity, and shared limits will not arise from collapse or wishful thinking. It demands design, experimentation, and institutional imagination—rooted in the best of human creativity, cooperation, and care.

How far could it go?

If expanded to include all life-sustaining, non-extractive, or regenerative sectors—even those currently commodified—Saito-style “care work” could plausibly encompass 50–60% of the U.S. economy, depending on interpretation. Key inclusions:

Healthcare & social assistance: 18% Education (public + private): 6% Food systems (farming, local distribution, cooking): ~5% Construction & housing (especially retrofitting, local/co-op): ~4% Renewable energy production & utilities: ~3–4% Local transportation & maintenance: ~2% Home goods & furnishing (durable, repairable): ~3% Disaster relief, emergency services, civil service: ~2–3% Arts, recreation, mental health: ~2%

These sectors shift from being capital-intensive to labor-rich, local, slow-growth, and sustainability-focused—more about use-value than exchange-value.

——-

Further Reading and References

1. Saito, Kohei – Slow Down (2024): Penguin
2. Hickel, Jason – Less is More (2020): https://jasonhickel.org
3. Raworth, Kate – Doughnut Economics (2017): https://doughnuteconomics.org
4. Graeber, David – Debt: The First 5,000 Years (2011): https://www.mhpbooks.com/books/debt/
5. Bookchin, Murray – The Next Revolution (2015): https://www.versobooks.com/products/1644-the-next-revolution
6. Ostrom, Elinor – Governing the Commons (1990): Cambridge
7. Bauwens, Michel – P2P Foundation: https://p2pfoundation.net
8. Manzini, Ezio – Design, When Everybody Designs (2015): MIT Press
9. Riofrancos, Thea – Resource Radicals (2020): Verso
10. Aronoff, Kate – Overheated (2021): https://www.boldtypebooks.com/titles/kate-aronoff/overheated/9781568589478/
11. Svampa, Maristella – Degrowth and South American Ecosocialism: https://degrowth.info
12. Open Source Ecology: https://opensourceecology.org
13. SlowLab & Cosmopolitan Localism: https://www.slowlab.net
14. Intercontinental Cry – Indigenous Governance: https://intercontinentalcry.org
15. Schneider, Nathan – Everything for Everyone (2018): https://nathanschneider.info/books/everything-for-everyone/
16. Creative Commons Movement: https://creativecommons.org
17. Commons Transition Network: https://commonstransition.org
18. Federici, Silvia – Re-enchanting the World (2018): PM Press
19. Decelerationist Manifesto (2013): PDF
20. Cosmolocalism & P2P Accounting: https://cosmolocal.org

all this via chatgpt (text, image), notebookLM (podcast, emoji), http://www.gan.ai (video avatar) and amol (questions, voice and person template)

Secretary Bessent:

China’s calculations:

The trade imbalance is three things not one:

• Goods (they net export a lot)
• Services (they net export a little)
• Multinational company revenues (they are way behind)

What if China thinks like Bessent?

• Match tariffs yesterday
• Sell Treasuries yields are up (not down)
  • Mess with the basis trade yesterday
• Seize US corporate assets in PRC
  • Cash in banks 
  • Business units: Apple, Microsoft, Tesla, Pfizer, GM, Intel, J&J, P&G, Starbucks, DuPont
  • Patents and trademarks. Clone everything
• Freeze select exports, “$1 parts that disable $50k products”
  • Block rare earth and lithium, raw and refined
  • Hold select commodity semiconductors
  • Lithium-ion batteries
• Block market access to Chinese buyers
  • China is the 3rd largest market for US goods
  • Soybeans, airplanes
• Plus the usual (American?) bag of tricks like deport, disappear and arrest (business) people randomly

WSJ itemizes: https://apple.news/A3hoE1j4ySpOPyYTqYEXflA

Key developments:

• October 6–10: Worst week in Dow history at the time (down ~18%). Global panic, frozen credit markets, and forced deleveraging hit all sectors.

• October 13: A massive one-day rally (+11% on the Dow) after coordinated G7 action and the U.S. Treasury’s plan to inject capital directly into banks.

• Ongoing fear: Despite policy moves (TARP, Fed liquidity programs), markets remained unstable due to continued bank failures, job losses, and earnings shocks.

Apple (US) makes $520 in profit per iPhone, paying its mostly US staff $12 per iPhone (for 15 minutes of work, i.e. all the HQ work divided by all the phones).

The top supplier companies (all Asia) make $41, and pay their people about $17 (for 2 hours of work).

If you brought the suppliers to produce locally in the US, you’d create an $8.50/hr job (McDonald’s pays more). Or if the job paid more, you’d raise the cost of the phone a bit.

But look more closely at what the supplier does. Sony, Samsung and TSMC are not just cheap labor. They have some unique technology and engineering. TSMC is an advanced manufacturer – so part of their ‘cost’ is HQ process engineering, and part of it is running their factory to spit out chips.

Here’s what TSMC looks like vs. GM (one estimate is ChatGPT and the other one is Grok):

Both companies produce about $1mm of revenue per employee, but look at how the employees are different. The engineering jobs at TSMC won’t come to the new Arizona factory they build, the factory worker jobs will. They don’t use very many of those; it’s only 60% of their cost (which above was about $2/chip).

What onshoring the supply chain to the US would look like is more like “offshoring” and “outsourcing” used to mean — high tech companies from Asia hiring uneducated, low-paid assemblers, call center reps, etc.

This is an image I found on Twitter, which doesn’t quite have it right.

That’s Richard Hofstadter, speaking in 1968, here.

And his most famous book; well, one of them.

Plus here are about 10 other snippets from 2007. (The “real” crash was Lehman, Sep 2008…). In them you see “crisis”, “next year will be worse”, “it could spread”, “the ratings agencies won’t fess up because they are earning fees”, the bailout will be hard because it will be politically unpopular…. All in 2007. Long before Sep 2008. In retrospect, all on George W Bush’s watch.

Busted deals

• IPOs pulled – eg Klarna
• M&A canceled – Q1 was a record high
• Equity and debt financings shelved
• New investments iced – factories
• Products canceled – eg Nintendo Switch 2

Cash preservation 

• Tax revenue falls
• Opex budgets cut – layoffs, hiring freezes
• Banks don’t lend or refi – shore up their lending portfolio because they expect failures 

Blowups

• Failed refinancings = bankruptcy 
• Failed fundraises = bankruptcy
• Margin calls on leveraged portfolios = fund blowups
• Banks get merged

If it’s a real recession that follows (it would given the above), then you get some “silver linings”

• interest rate cuts
• re-flood to stock markets

And you get some more problems

• incumbents raise first and best, getting stronger
• regulators come to fix the systemic risks uncovered “this time”
• elections swing 
• unemployment and general popular misery

https://en.wikipedia.org/wiki/List_of_stock_market_crashes_and_bear_markets

Charles Koch – lawsuit

Dan Loeb – deleted tweet “It was fun while it lasted”

Musk – said: hoping they are eliminated

Ben Shapiro, Daily Wire – “unconstitutional”

Galvin – bad rollout

Ackman – said: better to pause

JPM: “recession”

Polymarket: 66% chance of recession

Ted Cruz – it’s a tax

Sunday talk shows – BTC down 3% at 1130am ET

Futures for Monday as of Sunday —

Dow Jones Industrial average futures fell 1,405 points, or 3.7% Sunday evening, pointing to another brutal session ahead on Monday. S&P 500 futures shed 4.3%. Nasdaq-100 futures lost 5.4% as investors continued to shed their one-time tech winners to raise cash.

• Brown University: 23.9%

• Columbia University: 22.5%

• Cornell University: 15.6%

• Harvard University: 9.9%

• Princeton University: 9.6%

• Dartmouth College: 8.8%

• University of Pennsylvania: 17%

• Yale University: 12.2%

Under the agreement, Japan and other countries agreed to coordinate efforts to weaken the dollar by selling it in foreign exchange markets. This led to a sharp appreciation of the Japanese yen and German deutsche mark. While it temporarily helped reduce the U.S. trade deficit, the yen’s rapid rise also contributed to Japan’s asset bubble in the late 1980s, which later collapsed in the early 1990s, leading to Japan’s “Lost Decade.”

And regarding China…

Yes, there’s an informal parallel between Japan’s post-Plaza financial strategy and China’s massive purchases of U.S. Treasuries over the last few decades.

Key Comparison:

1. Japan in the Late 1980s: After the Plaza Accord forced yen appreciation, Japan faced slowing exports and an influx of capital. To sustain growth, it invested heavily in U.S. assets, particularly real estate and stocks, leading to a bubble that later burst.

2. China Since the 1990s: Instead of letting the yuan appreciate too quickly, China recycled its trade surplus by buying U.S. Treasuries, keeping the yuan stable and interest rates low in the U.S. (which fueled American consumer spending).

Outcome Differences:

• Japan’s bubble collapsed in the 1990s, leading to stagnation.

• China’s strategy avoided a Plaza-style shock, but U.S. reliance on China for debt financing created geopolitical and financial tensions, especially as China has reduced Treasury purchases in recent years to diversify its reserves.

In both cases, the U.S. trade imbalance led to financial interdependence, but China learned from Japan’s mistakes, choosing gradual financial shifts over abrupt currency revaluation.

Install https://github.com/open-webui/open-webui?tab=readme-ov-file — it’s two lines of command line

Next step: give it control of your browser — Install https://github.com/browser-use/browser-use

Here’s a concise punchlist to set up a new macOS machine with Ollama running DeepSeek on port 8080, exposed via ngrok, with the port change made permanent:

Punchlist for New Machine Setup

1. Install Ollama bash curl -fsSL https://ollama.com/install.sh | sh
2. Pull DeepSeek Model bash ollama pull deepseek-r1
3. Set OLLAMA_HOST Temporarily (Test) bash export OLLAMA_HOST="0.0.0.0:8080"
4. Start Ollama on Port 8080 bash ollama serve
5. Test Ollama Locally bash curl http://localhost:8080
6. Make OLLAMA_HOST Permanent (zsh) bash echo 'export OLLAMA_HOST="0.0.0.0:8080"' >> ~/.zshrc source ~/.zshrcOR for bash (if applicable): bash echo 'export OLLAMA_HOST="0.0.0.0:8080"' >> ~/.bash_profile source ~/.bash_profile
7. Verify Persistence bash echo $OLLAMA_HOST
8. Install Ngrok (if not installed) bash brew install ngrok/ngrok/ngrok
9. Set Ngrok Authtoken (replace YOUR_TOKEN) bash ngrok config add-authtoken YOUR_TOKEN
10. Start Ngrok Tunnel to 8080 bash ngrok http 8080
11. Test Ngrok Tunnel (replace with your ngrok URL) bash curl https://your-ngrok-url
12. Test DeepSeek via Ngrok (replace with your ngrok URL) bash curl https://your-ngrok-url/api/generate -d '{"model": "deepseek-r1", "prompt": "Test from ngrok"}'

Notes

• Replace YOUR_TOKEN with your ngrok authtoken.
• Replace your-ngrok-url with the forwarding URL from ngrok (e.g., https://abc123.ngrok.app).
• Use zsh commands unless echo $SHELL shows /bin/bash, then use bash alternatives.
• Run commands in order; skip installation steps if tools are already present.

At this point you can talk to Deepseek through Terminal. To create an app interface, you need to basically build a local HTML-engine app (Electron) because using a browser creates too many headaches with local security.

So you can do this. I would just copy and paste the below into your AI of choice and you’ll get all the spelled out how-to steps in better detail.

Steps to Create an Electron App for Ollama/Deepseek

1. Install Prerequisites:
   • Install Node.js and npm if not already installed.
2. Create Project Directory:bashCopyEditmkdir deepseek-chat cd deepseek-chat
3. Initialize Node.js Project:bashCopyEditnpm init -y
4. Install Electron:bashCopyEditnpm install electron --save-dev
5. Create Main Application File:
   • Create main.js:
   javascriptCopyEditconst { app, BrowserWindow } = require('electron'); function createWindow() { const win = new BrowserWindow({ width: 800, height: 600, webPreferences: { nodeIntegration: true, contextIsolation: false, }, }); win.loadFile('index.html'); } app.whenReady().then(createWindow);
6. Create HTML File:
   • Create index.html with a chat interface that interacts with the Deepseek API:
   htmlCopyEdit<!DOCTYPE html> <html lang="en"> <head> <meta charset="UTF-8"> <meta name="viewport" content="width=device-width, initial-scale=1.0"> <title>Deepseek Chat</title> <style> /* Add styles here */ </style> </head> <body> <div id="chat"></div> <input type="text" id="input" placeholder="Type your message here..." /> <button id="sendButton">Send</button> <div id="response"></div> <script> /* Add JavaScript for API calls and formatting */ </script> </body> </html>
7. Set Up API Call:
   • In the <script> section, implement the sendQuery function to send requests to your Ollama/Deepseek server via ngrok:
   javascriptCopyEditconst apiUrl = "https://0eaf-96-239-72-6.ngrok-free.app/api/generate"; async function sendQuery() { // Handle sending input to the API and displaying the response }
8. Update package.json:
   • Ensure the entry point is set to main.js:
   jsonCopyEdit"main": "main.js",
9. Add Start Script:
   • Modify the scripts section in package.json to include:
   jsonCopyEdit"scripts": { "start": "electron ." }
10. Run Your App:bashCopyEditnpm start
11. Format Output:
    • Improve the readability of responses by adding HTML formatting, CSS classes, and replacing newline characters with <br> tags.
12. Test and Debug:
    • Use the developer tools in Electron to debug and ensure the application works as expected. Monitor the Ollama server logs for incoming requests.

Final Notes

• Make sure your Ollama server is running and accessible via the ngrok URL.
• Adjust the API call in your Electron app to match the expected structure and endpoint of your Deepseek server.

Google, founded in 1995, reached >12bn revenue 12 years later, but it didn’t >2x the next year.

But perhaps the “End of Men” is a bigger force in reshaping politics than the “Immigrant Invasion”

This guy, pollster David Shor, has a very global POV on how sentiment worldwide is moving in politics

https://www.nytimes.com/audio/app/2025/03/18/opinion/ezra-klein-podcast-david-shor.html?referringSource=sharing

Saw similar forces in US, Germany, Norway… young people generally becoming conservative, and young men becoming ultra conservative and he speculates why

Eg “The End of Men” forces like from the Hannah Rosin book plus and this recent New Yorker piece on the same themes, “The Battle for the Bros”

https://apple.news/A2JiKJ-rwSkCcsZvMjYyyOw

Is not a hidden point. There is a whole Tucker Carlson franchise about it

Why Tucker Carlson’s ‘The End of Men’ Is So Disturbing – Men’…

I have felt it’s not a local US phenomenon, and Shor knows so. Whoever the candidate, whatever the local situation… young dudes are getting more conservative and even women too but maybe less so. 

Here is a glorious collection of charts, https://bsky.app/profile/lauraserra.bsky.social/post/3ll5lluhgek2f

And it made me remember what I really liked on Twitter and rarely find on LinkedIn are interesting, factual, surprising and debated ideas.

Well it happened for the first time today with this fascinating data drop thread.

https://bsky.app/profile/lauraserra.bsky.social/post/3ll5lluhgek2f

Elon is playing for 10x at this moment. Tesla down massively, reputation majorly contested, non-US SpaceX customers getting queered….

The thought was: this guy puts it all on the table to get to the next 10x.

The view was mostly respect.

This one has raised <500k and has 100 headcount, doubled in the last year. Does SaaS for car mechanic shops.

Still something about this picture kind of struck me.

Here are some other stars from Riyadh’s ecosystem

World War I (1914–1918)

At the time, Ireland was still part of the UK. Around 200,000 Irishmen served in the British Army, many believing Home Rule would follow. Others saw the war as irrelevant or an imperialist cause. Meanwhile, the 1916 Easter Rising, led by republicans opposed to British rule, marked a turning point. The British executed its leaders, fueling Irish nationalist sentiment and leading to the War of Independence (1919–1921).

World War II (1939–1945)

Ireland, now independent as the Irish Free State (later Éire), remained neutral. Taoiseach Éamon de Valera pursued strict neutrality, refusing to join the Allies despite pressure from Britain and the U.S. However, about 50,000 Irish citizens volunteered for British forces. Ireland provided weather reports crucial for D-Day but also maintained diplomatic ties with Germany, famously offering condolences for Hitler’s death, which Britain and the U.S. criticized.

Decline of the British Empire & Ireland’s Evolution

• 1922 – Anglo-Irish Treaty creates the Irish Free State; Northern Ireland remains under British rule.

• 1937 – Ireland adopts a new constitution, asserting greater sovereignty.

• 1949 – Ireland declares itself the Republic of Ireland, formally leaving the Commonwealth.

• 1950s–60s – Ireland stays out of Cold War alliances while Britain faces crises like Indian independence (1947) and Suez (1956).

• 1969–1998 – The Troubles erupt in Northern Ireland; Ireland diplomatically supports reunification but avoids direct involvement.

• 1998 – Good Friday Agreement stabilizes Northern Ireland, reducing tensions.

As Britain’s empire collapsed, Ireland moved from a post-colonial state to an independent European nation, eventually joining the EU (1973) and shifting from economic stagnation to rapid growth (Celtic Tiger era)

https://stephenfollows.com/p/why-are-actors-getting-older

But more importantly, here’s the WSJ talking it up as a longevity drug, because apparently people do think as I do that we have already found the longevity miracle drug.

Out of office

Recent story: A Harvard lab doing work on cancer won a grant in Q4. Expected the funds in February. Started firing up the experiments and hiring. Now it’s March. No checks. They called their grant administrator – out of office. Turns out they no longer work at the agency. Tried calling the boss. No response. So what happens next? They shut it down…

This is just one lab and the situation at every lab is likely identical. The Bloomberg piece, and other comments from contracts suggest the same – so many cases like this.

In recent years, the National Institutes of Health (NIH) has issued approximately 58,000 grants annually.  The National Science Foundation (NSF) awards about 11,000 new grants each year.  Regarding Small Business Innovation Research (SBIR) grants, across all federal agencies, approximately 4,000 awards are made annually.  Therefore, combining these figures, there are roughly 73,000 individual recipient projects funded each year by these programs.

One of those grants is probably for a rare disease one of your relatives has, sadly now kiboshed.

Indeed, you can browse awardees and research topics in these databases, all out in the open:

1. NIH RePORTER (reporter.nih.gov) – Search NIH-funded projects by year, institution, investigator, keywords, and funding mechanism.

2. NSF Award Search (nsf.gov/awardsearch) – Browse all NSF grants, including abstracts and funding details.

3. SBIR.gov Awards Database (sbir.gov/sbirsearch/award/all) – Search SBIR/STTR awards across all participating federal agencies.

But what about administrative overhead fraud?

Part of how this topic has been positioned in the political debate is that it’s actually about elite institutions, taking an unfair share of the grand amount for administrative overhead. The comments above are actually not about that. It’s about a total freeze on funds going out. Indiscriminate.

But let’s have a look at the administrative costs. Why would there be any administrative costs?

Shouldn’t you be able to do some biology research just by yourself with a couple hundred dollars worth of ingredients if you’re at a place like Stanford? Or is it a big deal that you need lab space, in particular lab space in a facility that surrounded by a bunch of other researchers who know what you’re working on, and in particular with other equipment that you might only use for five minutes, but that’s hugely expensive to get access to…?

(ChatGPT and me below.)

Some context on “indirect costs”

Indirect costs, also known as Facilities and Administrative (F&A) costs, cover expenses like utilities, maintenance, and administrative support essential for conducting research but not directly attributable to specific projects. Traditionally, institutions negotiated these rates individually with federal agencies, leading to variations based on each institution’s infrastructure and administrative needs.

The Small Business Innovation Research (SBIR) program, aimed at fostering innovation among small businesses, has historically capped indirect cost rates at 15% to ensure that a significant portion of funding is directed toward direct research activities. This cap reflects the program’s focus on maximizing research output while maintaining fiscal responsibility.

In February 2025, the National Institutes of Health (NIH) announced a policy to standardize the indirect cost rate at 15% across all grants, affecting both new and existing awards. This move aimed to streamline funding allocations and align NIH practices with those of private foundations, which often limit indirect costs. However, this decision faced criticism from research institutions, as it significantly reduced funding for essential operational expenses previously covered by higher negotiated rates. 

Institutions like Harvard University have historically negotiated higher indirect cost rates, sometimes exceeding 60%, to cover substantial investments in research infrastructure, compliance requirements, and administrative support. These negotiated rates reflect the actual costs incurred by large research institutions to maintain state-of-the-art facilities and support services necessary for cutting-edge research. 

Regarding the concentration of grant funding, data indicates that a significant portion of NIH funding is awarded to a relatively small number of institutions. For instance, in 2024, Johns Hopkins University led the rankings, with other top institutions also receiving substantial NIH funding.  This concentration underscores the pivotal role major research universities play in advancing scientific and medical research in the United States.

Would it help small institutions, who have fewer overhead costs?

Standardizing indirect cost rates at 15% could have mixed effects on smaller research institutions:

Potential Benefits:

• More Competitive Grants: Since large institutions like Harvard typically receive 50-60% indirect costs, reducing all institutions to 15% levels the playing field. Smaller universities might have a better chance of winning grants.

• More Funds for Direct Research: A lower cap means a larger portion of the grant must go toward actual research, which could benefit researchers at institutions with lower overhead costs.

Potential Downsides:

• Financial Strain on Small Institutions: Large universities often cross-subsidize research with other revenue sources. Smaller institutions, lacking these resources, may struggle to maintain compliance, infrastructure, and administrative support.

• Less Capacity to Scale Research: Without sufficient indirect cost recovery, smaller institutions may find it harder to expand lab facilities, hire support staff, or maintain cutting-edge equipment.

Why are the big ones expensive?

The high overhead (indirect) costs at major research universities come from a mix of necessary infrastructure, compliance burdens, and administrative expenses. Some are genuinely expensive and essential, while others may reflect inefficiencies.

Legitimate High-Cost Infrastructure

1. Cutting-Edge Research Facilities – Advanced labs for biomedical, physics, or AI research (e.g., clean rooms, supercomputing clusters, high-field MRI scanners).

2. Core Scientific Services – Genomics, mass spectrometry, high-performance computing, and other centralized resources available to researchers.

3. Regulatory Compliance – Staff and systems to manage IRB (human subjects research), IACUC (animal research), biosafety, export control, and federal grant reporting.

4. Library & Data Resources – Subscriptions to costly scientific journals, databases, and research tools.

5. Facility Maintenance & Utilities – Climate-controlled labs, 24/7 security for sensitive research, and high electricity costs for computing.

6. IT & Cybersecurity – Large institutions handle sensitive data, requiring secure infrastructure to meet NIH, NSF, and HIPAA compliance.

Potential Bureaucratic Bloat

1. Administrative Overhead – Universities have expanded non-teaching administrative roles significantly over the past few decades. Some of this is necessary, but studies suggest a high growth rate in non-research staff.

2. Overcompensated Leadership – Presidents, provosts, and senior administrators at elite universities often earn salaries in the millions.

3. Inefficient Decision-Making – Bureaucratic complexity can slow down research, with layers of approvals for simple tasks.

4. Luxury Campus Amenities – While not directly covered by grants, some universities spend heavily on non-essential perks (high-end fitness centers, fancy dining facilities), adding to operational costs.

Comparison to Smaller Institutions

Smaller institutions may lack certain rare, expensive facilities, but they also operate with leaner administrative structures. They often collaborate with larger universities or use shared national research facilities (e.g., national labs) instead of maintaining their own.

And do they produce more useful stuff?

Big universities are better at securing grants because of brand reputation, established networks, and lobbying power, but their efficiency in delivering high-impact research per dollar is debatable.

Why They Win More Grants

1. Reputation & Trust – NIH/NSF reviewers favor institutions with a strong track record.

2. Grant Writing Expertise – Large teams specialize in crafting winning proposals.

3. Faculty Prestige – Established researchers with prior funding get preference.

4. Political Influence – Elite universities have strong ties to policymakers.

5. Infrastructure Advantage – They already have the facilities, compliance systems, and personnel in place, making them “lower risk” for funding agencies.

Are They More Efficient?

Efficiency in High-End Research:

• Yes, in some cases – If a project requires multi-million-dollar facilities (e.g., particle accelerators, gene sequencing centers), large universities offer economies of scale.

• No, in other cases – Some research (especially theoretical or computational work) doesn’t require expensive infrastructure, yet still incurs high overhead at big institutions.

Inefficiencies in Bureaucracy:

• Administrative Bloat – The growth of university administrative staff outpaces research faculty, driving up costs.

• Slow Decision-Making – Large universities have complex layers of approval, making research slower and costlier.

• Salary Inflation – High executive pay and administrative costs can dilute funding impact.

Bottom Line

Big universities dominate funding because of structural advantages, not necessarily because they produce the most research per dollar. Smaller institutions might be more efficient in some areas but lack the infrastructure for certain types of research. If the goal is maximizing the impact of research dollars, a better funding model might involve targeted grants for smaller universities or requiring efficiency metrics in indirect cost negotiations.

So are the high costs just ignored?

Grant administrators do consider institutional overhead rates, but they generally do not adjust funding dollar-for-dollar based on whether a project is at a big vs. small institution.

How Indirect Costs Are Handled in Grant Funding

1. Fixed Indirect Cost Rates – Universities negotiate indirect cost rates with federal agencies, which then apply uniformly to all grants, regardless of the project’s actual administrative burden.

2. No Adjustments for Institution Size – A theoretical, low-capex project at a small school would have lower total costs because of a lower overhead rate, but big universities don’t receive less for the same project—agencies typically fund based on the approved budget, including the institution’s standard indirect costs.

3. Capped Overhead for Certain Grants – Some programs, like SBIR, cap overhead at 15%, but this applies universally and doesn’t scale by institution size.

Do High-Overhead Schools Adjust for Project Needs?

Not meaningfully. While universities justify their high indirect cost rates based on large-scale research needs, those rates still apply to all projects—even theoretical or computational work that doesn’t require expensive shared facilities.

In theory, universities could discount or waive indirect costs for projects that don’t require expensive infrastructure. In practice, they rarely do, since indirect cost recovery is a major revenue stream. Some elite universities have internal policies for reducing overhead on specific types of grants (e.g., certain private foundation grants), but federal grant overhead is generally non-negotiable at the institutional level.

Potential Reforms That Could Improve Efficiency

• Tiered Overhead Models – Adjusting indirect cost rates based on project type (e.g., theoretical vs. lab-based research).

• Performance-Based Overhead Scaling – Linking overhead rates to research productivity or efficiency metrics.

• Institution-Size Adjusted Grants – Providing additional incentives for smaller institutions with lower indirect cost burdens.

Right now, the system largely favors big universities charging high overhead, regardless of the project type

An example

Yes, your grant request from Hamilton College is likely to be smaller than an equivalent one from Yale because of differences in indirect cost rates.

Key Reasons Why

1. Lower Indirect Cost Rate at Hamilton – Smaller colleges typically have lower negotiated indirect cost rates (e.g., ~40% or less), while Yale’s could be 60%+, meaning a larger share of a Yale grant goes to overhead rather than direct research.

2. Lower Administrative Costs at Hamilton – A mathematics project at Hamilton likely requires minimal infrastructure beyond office space and computing, while Yale’s grant budget might still include institutional admin, compliance, and facility costs.

3. Same Direct Costs, Different Total Costs – If the research itself costs $100,000 in direct expenses:

• Hamilton (assuming 40% indirect): Total request = $140,000

• Yale (assuming 60% indirect): Total request = $160,000

Does This Make Hamilton More Competitive for Funding?

Not necessarily. Grant agencies typically fund the best proposals, not just the cheapest ones. If a panel prefers Yale’s reputation, established faculty, or research environment, they may fund the larger grant even if it’s more expensive. However, for smaller grants where cost-efficiency matters, Hamilton’s lower total cost might be an advantage.

Could Hamilton Voluntarily Charge Yale-Level Overhead?

No, because overhead rates are pre-negotiated with federal agencies and apply uniformly to all grants at a given institution. If Hamilton’s rate is 40%, it can’t arbitrarily raise it to Yale’s 60% level.

In summary: Yes, your grant request at Hamilton would likely be smaller, but that doesn’t necessarily make it more competitive unless the funding agency explicitly considers cost efficiency.

Why Reputation Often Beats Cost Efficiency

1. Peer Review Bias – Grant selection panels, often composed of researchers from top institutions, tend to favor proposals from well-established universities with stronger faculty track records.

2. Risk Aversion – Funders may see elite universities as a “safer bet” because they have a history of producing high-impact research and successfully managing large grants.

3. Network Effects – Researchers at top institutions often collaborate with other well-known scholars, making their proposals seem more promising.

But Don’t Administrators Have Fixed Budgets?

Yes, grant agencies work within a fixed budget, but they don’t allocate funds purely based on getting the most research per dollar spent. Instead, they prioritize:

• Scientific Merit (as judged by peer reviewers).

• Likelihood of Impact (often linked to institutional reputation).

• Track Record (past success with grants).

Does Being Cheaper Ever Help?

Sometimes, especially when agencies:

• Want to fund more projects – If they can stretch their budget by choosing lower-cost proposals, smaller schools might benefit.

• Target Underrepresented Institutions – Some NSF and NIH programs explicitly try to fund research at smaller colleges or minority-serving institutions.

Bottom Line

A smaller grant request alone won’t make you more competitive if reviewers see your institution or track record as weaker. Cost efficiency is rarely the deciding factor unless it aligns with a program’s funding strategy.

Other useful supplements:
– Arginine – vaso dilator
– Taurine – pre-GABA, anxiety reducer and mood stabilizer
– L-carnitine – energy production
– Creatine – only really important if you don’t eat a lot of meat or if you’re low on it (‘responder’); for all your cells

Some products out there:

One of the cooler things these days is they want to sell you islands. You’ve probably heard of some of their man-made epic formations like the Palm. Or the world.

Azeem has a solid post about the bifurcation in AI companies. Some use tons of capital (the AI Labs) and some use little (the Apps and Agents).

But how defensible are they?

Arguably the cash invested alone makes for a small set of AI Labs. The big ones are spending massively to train the next best model, and the infrastructure needed to do the training and inference. Plus lots of stellar engineers that need to invent ever better solutions that code autocomplete won’t solve for the next, next step. Or maybe not (DeepSeek, Manus AI.)

But probably rather defensible. OpenAI has been showing how. Stay a bit better, keep linking more pieces together, be full stack (are you renting the model or GPUs or data?), be the easiest way to get everything through one API (coding, pictures, agents, reasoning, voice, video.)

In Azeem’s top right are companies that are cheaper to get going but their advantage is stuff like: user relationships, brand, private or user-related data, and integrations into workflows.

If it works, OK. But are these really defensible as they seem? The next best agent will migrate your account in a flash, or even layer on top of your old account.

Say you want to attack finance and accounting. The UI and workflows, the bank records and invoices…historically it would be a pain to switch. Hence the old world has an SMB monopoly with Quickbooks and an enterprise monopoly with Netsuite.

But an agent can download every report, convert from PDF to local data structure, clone all the reports you want, and never ask you to use a UI again. You just ask for the forecast or the board deck. The finance agent makes all that lock in go away. Maybe except for brand.

And THAT agent has little such lock in itself. So why not switch immediately.

That’s the current state of consumer user AI chats.

It reminds me of the early search engines that all sucked equally – OpenText, Yahoo, Excite, Altavista, AskJeeves…

Eventually one of them was so much better, and built a self-reinforcing revenue and distribution wheel (AdWords and Adsense feed Chrome and iOS integrations which feed more user interaction data to improve search…) that they were the monopolist.

Data and Distribution, the lock-ins that made Google win. It probably can’t be brand loyalty.

Health Insurance in America : Throughline

https://www.npr.org/2025/02/27/1234145475/health-insurance-in-america

It pays with pieces of paper called Treasuries, that require their holders to get dollars, that make US currency very powerful. Plus other stuff that Balaji outlines in a few posts here, one of the first ones I’m seeing from Silicon Valley critics-of-the-left who are currently critical-of-the-right.

I guess only 2% of SpaceX employees have PhDs, though that’s 2-3x the average company.

Here’s Brian Leiter, who covers law schools and philosophy departments has been noting about how academia is faring lately.

Via ChatGPT

There are decades where nothing happens, and weeks where decades happen – and 2025 increasingly feels like the latter. Since Monday, a flurry of US tariff announcements, tariff delays, and reciprocal tariffs from allies unfolded at a dizzying speed. Disorienting is an understatement – but there are silver linings amidst the chaos, and an opportune moment to inject some optimism and insights I heard on the road this week.

I left our 10th annual Disruptive Tech Symposium in London a little tired, and a little jet lagged – but quite encouraged by a palpable energy, even amidst the geopolitical fragility. As of yesterday, European equities have outperformed other major markets globally – including the US – as investors quickly re-calibrate their expectations and allocation weightings for the region. Consider this eye-popping stat: in the last decade, 10x more capital has been invested in European technology than the prior decade (and more invested just last year than the prior decade combined). The conversations with founders and VCs underscored the enthusiasm around the pan-European opportunities in both TAM and technical talent. And while Brexit birthed a complicated perception of the European Union and a decade of cumbersome regulations, the pervasive feeling across the conference this year was one of solidarity. Even beyond the technology sector, that unity was galvanized by recent geopolitical events and the noteworthy news of Germany’s plans for fiscal expansion – a potential inflection point for the broader region.

It’s starting to feel like this pace of change and level of uncertainty will only persist – and regions and countries outside the US will be increasingly influential in defining this new era. It will behoove us all to think locally and globally along the next leg of this journey.

Though the quote may actually be a Mexican poet.

Though this quote is pretty great too.

Not quite what I expected to see.

Chaos makes prediction harder when:

1. The system is highly nonlinear and sensitive to small changes.

2. There’s a lack of sufficient training data.

3. The system undergoes structural breaks (e.g., black swan events).

AI is good at detecting statistical regularities but less effective when those regularities don’t hold over time.

Via ChatGPT

Naval Power Projection

• U.S.: 80% – 11 aircraft carriers, 68 submarines (including 14 nuclear ballistic missile submarines, SSBNs), 90 destroyers, and an extensive global naval base network (Japan, Guam, Bahrain, Diego Garcia, Rota, Norfolk).

• China: 15% – 3 aircraft carriers, 79 submarines (including 6 SSBNs), 50 destroyers, growing blue-water navy but lacks overseas bases and carrier battle group experience.

• Russia: 5% – 1 aging aircraft carrier (Admiral Kuznetsov, often non-operational), 58 submarines (11 SSBNs) but mainly focused on coastal defense and Arctic dominance.

Naval Theater Comparison

Region U.S. Dominance China Russia

Indo-Pacific 70% 30% 0%

Middle East 85% 10% 5%

Arctic/North Atlantic 50% 0% 50%

Africa/Indian Ocean 60% 30% 10%

Air Power Projection

• U.S.: 85% – 5,200+ combat aircraft (F-35s, F-22s, F-15s), 800+ tankers for global reach, thousands of UAVs (Predator, Reaper, Global Hawk), bases in 70+ countries.

• China: 10% – 2,200+ combat aircraft (J-20s, J-16s, but limited aerial refueling), lacks strategic bombers.

• Russia: 5% – 1,500+ combat aircraft (Su-57s, Su-35s, Tu-160 bombers), but relies on regional operations and lacks tankers for global power projection.

Air Power Theater Comparison

Region U.S. Dominance China Russia

Indo-Pacific 75% 25% 0%

Middle East 80% 5% 15%

Eastern Europe 65% 0% 35%

Arctic/North Atlantic 60% 0% 40%

Ground Forces & Global Deployment

• U.S.: 70% – Army: 1.3M active troops, 700+ overseas bases, rapid deployment (82nd Airborne, Marine Expeditionary Units).

• China: 20% – PLA: 2M troops, heavy land force but limited overseas deployment.

• Russia: 10% – 900K troops, strong regional forces but lacks global logistics.

Ground Forces Theater Comparison

Region U.S. Dominance China Russia

Indo-Pacific 60% 40% 0%

Middle East 70% 10% 20%

Eastern Europe 55% 0% 45%

Africa 50% 40% 10%

Final Global Military Power Projection Estimate

Domain U.S. China Russia

Naval Power 80% 15% 5%

Air Power 85% 10% 5%

Ground Forces 70% 20% 10%

Strategic Bases & Allies 90% 5% 5%

Key Takeaways:

• U.S. maintains global dominance in naval, air, and strategic base networks.

• China is rapidly growing but lacks overseas bases and blue-water naval experience.

• Russia is regionally powerful (Europe, Arctic, Syria) but lacks global power projection.

Adding Europe

Global Military Power Projection: U.S. vs. China vs. Russia vs. EU

The EU (primarily France, Germany, Italy, and the UK) has strong regional forces but lacks the global reach of the U.S. Here’s how they compare across naval, air, and ground power.

Naval Power Projection

• U.S.: 80% – 11 aircraft carriers, global naval bases, most powerful blue-water navy.

• China: 15% – 3 aircraft carriers, expanding navy but limited global presence.

• Russia: 5% – 1 unreliable aircraft carrier, strong Arctic fleet but lacks global reach.

• EU: 20% – 4 carriers (France’s Charles de Gaulle, UK’s Queen Elizabeth-class), 65+ submarines, strong Mediterranean and North Atlantic presence.

Naval Theater Comparison

Region U.S. EU China Russia

Indo-Pacific 70% 5% 25% 0%

Middle East 85% 15% 5% 5%

Arctic/North Atlantic 50% 40% 0% 50%

Africa/Indian Ocean 60% 30% 30% 10%

Air Power Projection

• U.S.: 85% – 5,200+ combat aircraft, 800+ tankers, global deployment.

• China: 10% – 2,200+ combat aircraft, growing but lacks refueling and strategic bombers.

• Russia: 5% – 1,500+ combat aircraft, strong regional but weak global reach.

• EU: 25% – 2,000+ modern fighters (Eurofighter Typhoon, Rafale, F-35), strong airlift but fragmented command.

Air Power Theater Comparison

Region U.S. EU China Russia

Indo-Pacific 75% 5% 25% 0%

Middle East 80% 30% 5% 15%

Eastern Europe 65% 40% 0% 35%

Arctic/North Atlantic 60% 40% 0% 40%

Ground Forces & Deployment

• U.S.: 70% – 1.3M troops, global bases, rapid deployment.

• China: 20% – 2M troops, large but lacks overseas deployments.

• Russia: 10% – 900K troops, focused on Europe and regional dominance.

• EU: 30% – 1.4M troops, strong NATO presence but lacks unified command.

Ground Forces Theater Comparison

Region U.S. EU China Russia

Indo-Pacific 60% 5% 40% 0%

Middle East 70% 40% 10% 20%

Eastern Europe 55% 50% 0% 45%

Africa 50% 40% 40% 10%

Strategic Base & Alliance Influence

Metric U.S. EU China Russia

Global Military Bases 90% 30% 5% 5%

Allies & Defense Agreements 100% 85% 30% 10%

Global Logistics & Mobility 90% 50% 15% 10%

Final Global Military Power Estimate

Domain U.S. EU China Russia

Naval Power 80% 20% 15% 5%

Air Power 85% 25% 10% 5%

Ground Forces 70% 30% 20% 10%

Strategic Influence 90% 50% 5% 5%

Key Takeaways

1. U.S. remains dominant globally, with unmatched force projection, bases, and alliances.

2. EU is a significant power but lacks strategic unity and global force projection.

3. China is growing but still limited by its lack of allies and global bases.

4. Russia is regionally strong in Eastern Europe and the Arctic but lacks global reach.

An assessment of this view

# Analysis of U.S., Russia, and China Military Power Comparison

Your comparative analysis provides a structured overview of military capabilities, but there are several areas where it warrants refinement for greater accuracy and context.

## Strengths of the Analysis

– The theater-by-theater breakdown is analytically useful

– The comparative approach across domains (naval, air, ground) provides useful structure

– Including EU powers adds important context to the global balance

## Areas Requiring Reconsideration

### Percentage Allocations

The percentage-based approach creates an impression of precision that may be misleading. Military power is multidimensional and context-dependent. For instance:

– The 80% U.S. naval dominance figure oversimplifies complex factors like operational experience, geographical advantages, and technological edges

– Theater-specific percentages don’t fully capture how forces would actually perform in conflict scenarios

### Naval Power Assessment

– **Russia**: The assessment undervalues Russia’s submarine capabilities and Northern Fleet strength

– **China**: The 15% figure understates China’s rapid naval expansion, particularly in the South China Sea

– **U.S.**: While U.S. carrier superiority is significant, the vulnerability of carriers to modern anti-ship missiles isn’t addressed

### Air Power Considerations

– Modern air defense systems (like Russia’s S-400/S-500) significantly complicate air superiority calculations

– Quality differences between aircraft generations matter more than raw numbers

– The analysis doesn’t account for stealth capabilities and electronic warfare advancements

### Ground Forces Reality

– The effectiveness of ground forces depends heavily on logistics, which isn’t fully reflected

– Modern conflicts demonstrate that numerical superiority doesn’t translate directly to battlefield success

– Special operations capabilities are not adequately addressed

## The Nature of the Rivalry

The military rivalry between these powers is multifaceted:

– **Asymmetric Approaches**: Russia and China focus on countering U.S. advantages rather than matching them symmetrically

– **Hybrid Warfare**: The rivalry extends beyond conventional forces to cyber operations, information warfare, and proxy conflicts

– **Regional vs. Global**: China and Russia prioritize regional dominance in their respective spheres of influence, while the U.S. maintains global commitments

## Additional Considerations

1. **Nuclear Deterrence**: Strategic nuclear forces fundamentally alter the calculus of direct confrontation

2. **Defense Industrial Base**: Sustainability of military operations depends on industrial capacity

3. **Technological Edge**: Investments in AI, hypersonic weapons, and space capabilities are reshaping traditional power balances

4. **Economic Constraints**: Military spending as a percentage of GDP varies significantly among these powers

5. **Political Will**: Domestic support for military operations significantly impacts actual power projection

## Conclusion

While the U.S. maintains significant advantages in global power projection, the rivalry is more nuanced than percentage allocations suggest. Russia and China have developed capabilities specifically designed to counter U.S. strengths, particularly in their regional spheres of influence. The true balance of power would be highly dependent on specific conflict scenarios, geographical contexts, and the nature of political objectives.

<questions by me, answers by ChatGPT and anthropic>

These recommendations are close-ish — they do recommend a coffee shop, but they are wrong: not close, not quiet.

I think the “near” thing is the most egregious error though. It’s basically not thinking about that aspect. Kind of the way the early models were useless at arithmetic. ChatGPT and Gemini only seem to have understood the city I have in mind.

https://www.visualcapitalist.com/u-s-share-of-global-economy-over-time/

Hi Amol

XX kindly forwarded on your details. By way of introduction, I sit in the Innovative XX team for Global XX which focuses on innovations in the XX space.

A central vendor intake process has been implemented to enhance efficiencies within our team. Please complete the vendor intake form at the following  link – any additional supporting information can be sent to xyz@email. These details will be evaluated by a member of our team, and an invitation to participate in a vendor demo day will be offered if your services align with our business needs and objectives.

Hope to hear from you soon!

XX Person

XX Lead, Innovative XX Director

Business XX, XX and XX (BXXX)

XX Innovative XX

That sounds like innovation at the speed of light to me.

However it’s one step closer to a pill for longevity that’s actually cleared by FDA…albeit for dogs. There are none with that label for humans, though some folks are trying to get a trial OK’d for that purpose, it’s not historically a matter that FDA entertains. “Aging” isn’t a disease to them; only specific diseases can be.

There are other super cool approaches in the works for dogs and cats (and people) — hope I can post more about those soon.

Depression

Addiction – nicotine, alcohol, opioids

Shopping, gambling

Dementia and Alzheimer’s?

Crime?

And this micro-trend of microdosing is interesting

Twice as good as naltroxene (the actual drug for cutting addiction)

The topic: are people just LLMs? Hinton seems to think so.

https://podcasts.apple.com/us/podcast/stanford-psychology-podcast/id1611926356?i=1000691921952

Artificial Anthropology – I hope someone is documenting these early days

Artificial Sociology

Philosophy is going to have lots of good ethics stuff to do, and more on designing and comparing kinds of minds.

Psychology as a department is only 100 years old or so, and CS is maybe 60. The business schools are all about 10@ years old. Same for the engineering schools.

When there is a big change, there are usually whole new fields of study.

This is the product under the hood:

https://community.com/

The query was just “taylor swift top literary works”

32GB — after about 17 hours, here’s what it mustered for the query —

7b did a lot better, or at least faster. It took 1 min. But the answer is horrible. Lots of nonsense in there. There is not a memoir, and there are no 1990s albums, or songs called “we R U overpriced” (literally a rare zero-results google search).

Of course, in only 2 seconds the 1.5b model spit this out, and it’s all made up – a great example of hallucinations. But it gets something subtly right. Some subject matters are the kinds of things her songs are about (though not race…):

Germany- aligns with China on trade and military arrangements.

France- warms to Russia. Opens visas and business.

Those two changes weakens the EU.

Japan- flips to start relations with China. And Korea.

Those three weaken the US position, for example:

– end dates on US bases in Germany and Japan and Korea

(Saves the US money? Expands the influence of others? And what if Russia and China put bases in those places?)

Here’s another vector:

Haiti- deal with Nigeria or Russia or France. Military bases 60 minutes from Florida.

Mexico- collab with China. Why not when the US is all negativity? Bases? Joint air command?

Just some ideas. But worth considering the opposite of the world order. Germany’s hard right is <50% of the vote right now – perhaps 30%? – what would it take for American policy to flip it?

Europe actually beat China. That’s news.

Enterprise software, aka “generalist AI + other stuff” was the obvious number 1. But note yet again how far Health is ahead of Fintech (true for the last few years – and probably forevermore.)

These layers are nice to look at. The red and green are where interest has been pivoting back and forth. The “DeepSeek World” would argue for more green. The “Titan World” (more on that soon) would argue for more red.

Deal Sourcing

1. Develop a framework for identifying high-potential startups in emerging industries using AI-driven tools.
2. Create a checklist for evaluating inbound pitch decks to quickly assess viability and alignment with fund focus. Plus a tool for processing the inbound flow of decks and email (e.g. export to NotebookLM using Google App Scripts)
3. Draft a script for cold outreach to founders of promising startups, emphasizing mutual value.
4. Design a system for tracking and prioritizing warm introductions from your network. Plus build the “do it” part (e.g. Bardeem or your own browser-integrated local LLM.)
5. Develop a strategy for leveraging LinkedIn to identify and connect with founders in specific sectors.
6. Create a guide for attending startup events and conferences to maximize deal flow opportunities.
7. Draft a template for a follow-up email after meeting a founder at a networking event.
8. Develop a process for analyzing Crunchbase or PitchBook data to identify startups raising soon.
9. Create a playbook for collaborating with accelerators and incubators to source deals.
10. Design a scoring system for evaluating startups based on team, market, traction, and technology.

Due Diligence

11. Develop a comprehensive due diligence checklist for evaluating seed-stage startups.
12. Create a framework for analyzing a startup’s competitive landscape and market positioning. First draft of this is exceedingly well done by generalist LLMs.
13. Draft a guide for conducting reference calls with a founder’s previous colleagues or investors. Plus how about a tool for recording the actual meeting and summarizing (there are infinitely many.)
14. Design a process for evaluating a startup’s financial projections and unit economics.
15. Develop a template for summarizing key findings from technical due diligence.
16. Create a system for tracking and comparing term sheets from competing VC firms. Comparing legal docs for differences, a form of thoughtful redlining, is also default and great.
17. Draft a script for asking insightful questions during founder interviews.
18. Develop a methodology for assessing a startup’s scalability and potential exit opportunities.
19. Create a guide for evaluating the strength and defensibility of a startup’s intellectual property.
20. Design a framework for assessing a startup’s cultural fit with your fund’s portfolio.

Portfolio Support

21. Develop a strategy for helping portfolio companies with hiring key executives.
22. Create a template for quarterly portfolio updates, including KPIs and milestones.
23. Draft a guide for facilitating introductions between portfolio companies and potential customers.
24. Design a system for tracking and reporting on portfolio company performance to LPs.
25. Develop a playbook for assisting portfolio companies with follow-on fundraising.
26. Create a framework for organizing portfolio-wide events to foster collaboration.
27. Draft a script for delivering constructive feedback to founders during board meetings.
28. Develop a methodology for identifying synergies between portfolio companies.
29. Create a guide for helping portfolio companies navigate pivots or strategic shifts.
30. Design a process for conducting post-mortem analyses on underperforming investments.

Bonus: Personal Branding and Thought Leadership Prompts

1. Content Creation
   • Develop a content calendar for sharing insights on emerging trends, startup advice, and investment theses on LinkedIn and Twitter.
   • Draft a blog post template for analyzing industry trends and their implications for startups and investors.
2. Networking and Visibility
   • Create a strategy for building relationships with influential founders, investors, and journalists in the tech ecosystem.
   • Draft a script for reaching out to podcast hosts or event organizers to secure speaking opportunities.
3. Social Media Strategy
   • Design a framework for consistently engaging with your audience on LinkedIn (e.g., commenting on posts, sharing insights, and posting original content).
   • Develop a guide for using Twitter to share quick takes on startup news, funding rounds, and market trends.
4. Thought Leadership
   • Write a template for publishing a monthly newsletter summarizing key insights from your deal flow, portfolio updates, and industry observations.
   • Create a framework for hosting webinars or AMAs (Ask Me Anything) to share your expertise with founders and aspiring VCs.
5. Personal Branding
   • Develop a personal branding statement that clearly communicates your unique value proposition as a VC Associate.
   • Draft a guide for creating a personal website or portfolio showcasing your investment thesis, portfolio contributions, and thought leadership content.
6. Public Speaking
   • Create a checklist for preparing and delivering impactful presentations at industry events or panels.
   • Draft a pitch for why you should be invited to speak at a specific conference or event.
7. Writing and Publishing
   • Develop a process for co-authoring articles with portfolio founders or other investors to amplify your reach.
   • Draft a template for submitting op-eds or guest articles to tech publications like TechCrunch, Forbes, or VentureBeat.
8. Community Building
   • Design a strategy for building and engaging with a community of founders, operators, and investors (e.g., Slack groups, Twitter communities, or local meetups).
   • Create a guide for hosting virtual or in-person events to foster connections within your network.
9. Mentorship and Giving Back
   • Draft a script for offering mentorship to aspiring VCs or early-stage founders.
   • Develop a framework for sharing lessons learned from your VC journey through mentorship programs or online forums.
10. Metrics and Tracking
    • Create a system for tracking the impact of your thought leadership efforts (e.g., engagement metrics, follower growth, or inbound deal flow).
    • Draft a template for setting and reviewing quarterly goals for personal brand growth.

40%

Sad to see the OG is offline though. Bring it back!

https://setiathome.berkeley.edu/

https://slate.com/news-and-politics/2024/04/donald-trump-news-2024-weight-loss-ozempic.html

Consider:

Shout out to Mitchell Hauser

True in all things (making companies, medicines…) since you need time to figure it out but also time to ship it in perfect time for everyone else to understand.

A neat example from far left field is “Cat person”, the most read and viral short story that the New Yorker ever published, here.

It appeared December 2017, a few months after MeToo went nuclear. And it had a major insight – it’s about non-MeToo relationship / sex awkwardness. It points out the ambiguity, the difficulty of judging in a gray area, and certainly lands with the relatability of “I felt that too.”

But it got written in April, pitched over the summer, read perhaps after MeToo September, and run in December.

So that artist had been meditating on the signals of the universe – personal experiences that Spring, Trump’s Access Hollywood tape that prior fall (as she says.)

Back to Rick Rubin – he’s kind of a Buddhist koan writer now and the book is a nice collection of inspiring but familiar statements. Especially the “Breaking the Sameness” tips – a rare section where you are literally getting to housefly Rubin in the studio back in those moments of creation (JayZ 99 Problems? Chili Peppers Under the Bridge?)

Deep cuts: Brian Eno’s playing cards. Check those out to find inspiration.

It says something like “get our safe drug that’s cheaper and lose weight” but without any caveats or explanations or warnings.

Maybe all those warnings are pointless….

Or maybe they clatter about and add up, in the way people say things like “I don’t want that stuff in my body.” That happens with stuff like vaccines right?

Here’s the ad they ran (compare it to any other ad for a medicine you have ever seen):

Last night at dinner in Dubai with the head of a major British agency, we went through the math.

It’s actually a lot easier than this, from Musk:

In fact you can just do it off the top of your head like this:

Use GDP, is $15T, and the US govt share of that is around 1/3, so the US Federal budget is $5T about.

It breaks down like this:
$1T Medicare
$1T Social Security
$1T Medicaid
$1T Military
the other 1ish T is other stuff, of which interest is variable and lately very high

Compare what we came up with over dinner with the chart above. Roughly accurate.

Perhaps you’ll be surprised that the front page of the news for a week has been USAID, which is $70bn or so per year, or about 1%.

Most folks talk about the intractability of changing the curve….but, note this: $2T or 40% is health care costs. Canada spends about half per person. What if the US did? That’s $1T right there.

But forget the Canada and international politics matters — here’s the way to cut the healthcare budgets.

AI to assist scientists, doctors, administrator, insurers and patients, obviously. My McKinsey friend Rodney Zemmel and Cutler from Harvard + co-authors here predict 5-10% savings in 5 years from today’s right-now tech: that’s $100-200B of the Medicare/Medicaid problem above.

Here’s the paper at NBER

The singularly most powerful application of AI is where labor costs are high — and health care is 1/2 labor. Of the $4.5T spent on healthcare in the US in 2022, here’s the breakdown

1. Labor (~50% | $2.2T)

• Physicians & Nurses (~20% | $900B) – Salaries, benefits, admin burden
• Other Health Workers (~15% | $675B) – Technicians, aides, therapists
• Administrative Staff (~10% | $450B) – Billing, insurance processing
• Non-Clinical Staff (~5% | $225B) – Hospital support, cleaning, food service

2. Consumable Opex (~30% | $1.3T)

• Drugs & Medical Devices (~15% | $675B) – Prescription drugs, implants, machines
• Medical Supplies (~5% | $225B) – Bandages, syringes, PPE
• Diagnostics & Labs (~5% | $225B) – Imaging, blood tests
• Insurance Overhead & Profits (~5% | $225B) – Admin costs, margins

3. Real Estate & Infrastructure (~10% | $450B)

• Hospital Construction & Maintenance (~6% | $270B) – Depreciation, facilities
• Equipment Depreciation (~4% | $180B) – MRI machines, surgical tools

4. Other (~10% | $450B)

• Long-term Care & Home Health (~5% | $225B) – Nursing homes, home aides
• Public Health, Research, Misc (~5% | $225B) – CDC, NIH, health IT

sources

But some of our LifeX founders have been pointing out other payments that might abruptly cease — NIH… NSF…

Last week the websites for various health finance services were down for 24 hours. People got nervous that their hospitals wouldn’t get paid what is owed by the federal govt. But I’m not aware of folks that stopped giving care.

But if a science trial gets “cut” the scientists are not a business. They don’t have any other funds to keep the lights on.

And when a trial is interrupted – like the mice don’t get the injection one day, or the nuclear fissile material stops getting processed… you lose the whole study.

I am told that World Health Organization has told all its global grantees to EXPECT this to happen and stop research that might be broken by an interruption in funding.

Chaos good or chaos bad…?

From here https://llm-stats.com/models/compare/deepseek-r1-vs-o1-2024-12-17

But 250,000 people came last year. (WebSummit is 70k.)

That makes it the largest tech event in the world.

Though I don’t quite think it is the *best* for being biggest. Over the last decade, the Saudis have been copy-and-pasting Davos, CES, Tokyo, SpaceX, etc into the desert.

But still no beers.

Instead just consider if the hugely profitable business of Lilly and Novo could be just as profitable but deliver the drugs for $30/month instead of $1,000/month (they are already $250/month effectively when negotiated – e.g. in Europe or some US plans.)

Some math with some made up numbers sprinkled in for napkin purposes.

Baseline Scenario

• Current Adoption: 2% of US adults (~5.2M people)

• Price per Year: $10,000

• Manufacturer Margin: 90%

• Revenue: $52B

• Profit: $46.8B

We now calculate price points at 5%, 10%, 30%, and 50% adoption while keeping total profit constant at $46.8B.

Breakpoints for Equivalent Profit Levels

At 50% adoption, GLP-1 drugs could cost as low as $1,360 per year while maintaining current profits.

mpact of Cost Reductions

Higher production scales can reduce manufacturing costs, allowing even lower prices. If production costs fall from $1,000 to $500 per year, pricing could drop further:

If production costs shrink to $500 per person, a 50% adoption scenario could sustain the same profit at just $860 per year per patient.

Examples of Drug Cost Reductions Due to Scale

1. HIV Antiretrovirals (ARVs)

• Early price (~2000s): ~$10,000 per patient per year

• Later price (~2020s, generic competition & scale-up): ~$75 per year in low-income countries

• Key drivers: Global demand (PEPFAR, WHO programs), generic manufacturing, economies of scale

2. Statins (Lipitor, Crestor, etc.)

• Early price (~2000s, branded): ~$3,000/year

• Later price (post-generic competition, ~2020s): ~$50/year

• Key drivers: Patent expiration, mass production, generics driving cost down

3. PCSK9 Inhibitors (Cholesterol drugs, e.g., Repatha, Praluent)

• Early price (~2015, launch): ~$14,000/year

• Later price (~2020s): ~$5,850/year

• Key drivers: Competitive pricing, expanded manufacturing

4. Insulin

• Early price (~1990s-2000s): ~$5,000/year in the US

• Later price (~2023, capped generics): ~$500/year

• Key drivers: Policy pressure, biosimilars, efficiency in manufacturing

See? Doable.

And that’s not counting the impact of the 100+ similar drugs in pipeline that are not owned by Lilly/Novo.

But societies are big, complex spaces with lots of people. So politics is as important as economics. Or rather, how the pie is divided, who wins and loses, and differential outcomes seem to matter a lot — politics.

So there’s little doubt that reducing regulation (Millei 2025 in Argentina, Trump 2, Reagan, Thatcher, Chile’s Chicago Boys, Deng’s Chinese Capitalism, Singh’s post-Duty Raj India) spurs economies.

But the important question is why it comes and goes. Here are the arguments.

Strengths

1. Rapid Economic Growth – Robert Barro (1996)

• Paper: “Determinants of Economic Growth: A Cross-Country Empirical Study”

• Concept: Growth is higher in economies with lower government consumption and freer markets.

• Example: West Germany’s Wirtschaftswunder under Ludwig Erhard.

2. Innovation & Entrepreneurship – Joseph Schumpeter (1942)

• Book: “Capitalism, Socialism, and Democracy”

• Concept: Creative destruction—market competition fuels innovation.

• Example: Hong Kong’s deregulated, low-tax environment fostered business dynamism.

3. Fiscal Discipline – James Buchanan (1962)

• Book: “The Calculus of Consent” (with Gordon Tullock)

• Concept: Governments tend to overspend unless constrained; markets enforce discipline.

• Example: New Zealand’s Rogernomics cut deficits and reined in government debt.

4. Resilience Post-Crisis – Milton Friedman (1962)

• Book: “Capitalism and Freedom”

• Concept: Free-market economies recover faster from crises than planned economies.

• Example: Estonia’s rapid growth after Soviet collapse due to flat tax and deregulation.

Pitfalls & Perils

1. Rising Inequality – Thomas Piketty (2014)

• Book: “Capital in the Twenty-First Century”

• Concept: Capital accumulation in free markets leads to growing inequality unless taxed.

• Example: Chile’s inequality grew after privatization, leading to protests in the 2010s.

2. Social Unrest – Albert O. Hirschman (1970)

• Book: “Exit, Voice, and Loyalty”

• Concept: When economic policies create hardship, people either protest (voice) or disengage (exit).

• Example: Thatcher’s privatization policies led to widespread labor strikes.

3. Vulnerability to Crisis – Hyman Minsky (1986)

• Book: “Stabilizing an Unstable Economy”

• Concept: Deregulated financial systems are prone to speculative bubbles and crises.

• Example: Chile’s 1982 banking crisis resulted from financial deregulation.

4. Political Instability – Dani Rodrik (2007)

• Book: “One Economics, Many Recipes”

• Concept: Market reforms require political legitimacy, or they will be reversed.

• Example: West Germany’s ordoliberal model was softened under SPD rule in the 1970s.

5. Capture by Elites – Mancur Olson (1982)

• Book: “The Rise and Decline of Nations”

• Concept: Free markets often lead to rent-seeking, where elites entrench themselves.

• Example: Russia’s post-Soviet privatization led to oligarchic wealth concentration.

https://podcasts.apple.com/us/podcast/innovation-and-leadership-with-jess-larsen/id1099241776?i=1000625483219

• Ketone energy (from fat not carbs or protein)
• How it metabolizes in heart (better function), brain (clearer thinking), better lifespans (in mice)
• Why too much bad fat and protein are not ways to do it
• What a good way might be (BHB, “good fat” as your calories), which is a way to do it without long fasting (have calories from BHBs available easily vs after processing all bodily resources)

Plus some pointers to animal diets as a benchmark (not the “caveman” ultra-high protein)

Greg is the founder of Juvenescence, Biohaven and more – and a former ER doc before becoming an entrepreneur.

Vijay Pande from A16Z – an AI doctor on your phone who’s with you 24/7. Talking to you from the moment you wake up and monitoring everything all the time. — I think that sounds about right.

Zack from Jazz predicted: comfortable clothes and a the return of fat fashion (it’s uncool to be skinny in 2030.) — Maybe!

Others predicted more testing, diagnostics, and supplements.

My own handful of predictions (let’s see how they hold up) —

By 2030: 5 years, 5 blockbusters. Each as significant as the GLP1s. Starting with pain, this year. The new nerve drug for pain makes opioids obsolete and will create much other transformation. 4 more to go…

Longevity 2030 – it has been flat for a bit, but I predict measurable longevity results globally by 2030. Because of what already exists. GLP1s. Obesity is already falling in the US. Costs are falling. Penetration is rising. Obesity is what makes heart attacks and strokes happen (25% of mortality.)

Cell and gene therapies. 50% of people will have a gene therapy by 2030 – of some kind, perhaps wrinkles or hair color, but something. (I might be off by 5-10x, but today’s number is <100th of 5%.)

Drugs reborn; bad ==> good. Bad behaviors get defanged. Eg alcohol or drugs that don’t come with harms. Already happening really, if you check the gummy counter at most California shops.

Good luck buying a new house, or investing in new construction.

But tells you that AI without Ego can beat people with ego.

And USAID.gov…

But then the first query I ran went for 10 minutes and never completed.

https://www.biopharmaboardroom.com/news/11/2108/fda-grants-priority-review-to-vertexs-suzetrigine-targeting-january-2025-for-action-date.html

It blocks nerve signals for pain. It’s a pill.

The implications are going to be wide and varied.

Here’s my post from a couple days ago about why DeepSeek was cheap to make – because it used select datasets and OTHER MODELS.

OpenAI agrees. It was trained on OpenAI. The way OpenAI was trained on the New York Times…?

Here’s a good post by a human that kind of says what you see below, and below is written by a few iterations of questions to DeepSeek.

• Reinforcement Learning Focus: DeepSeek-R1-Zero uses pure reinforcement learning (RL) without supervised fine-tuning (SFT), reducing the need for costly labeled data and computation.
• Efficient Training Pipeline: DeepSeek-R1 employs a multi-stage process, including fine-tuning with high-quality reasoning data and RL stages for reasoning, safety, and usefulness, minimizing wasted computation.
• Knowledge Transfer via Distillation: DeepSeek transfers knowledge from larger models (e.g., Llama, Qwen) to smaller ones during distillation, reducing reliance on expensive base models for core training.
• High-Quality Data Curation: Focus on curated, reasoning-focused data allows DeepSeek to achieve strong performance without massive datasets, lowering data collection and processing costs.
• Flexible Base Models: DeepSeek can use various base models for distillation, offering cost savings if cheaper alternatives are available.
• Cost Savings in Data: Reduced reliance on large datasets decreases data acquisition and processing expenses.
• Cost Savings in Computation: Efficient training pipelines and RL reduce computational requirements.
• Cost Savings in Deployment: Smaller distilled models are cheaper to deploy and maintain.
• No Dependency on Llama for Core Training: While Llama models were used in distillation, DeepSeek’s core training relies on RL and curated data, not specific base models like Llama.
• Reduced Need for Labeled Data: DeepSeek’s RL approach eliminates the need for extensive labeled datasets, making it more cost-efficient compared to models like Llama that rely on supervised fine-tuning.

More detail

DeepSeek’s lower training costs compared to models like Llama or OpenAI’s are due to several smart strategies:

1. Reinforcement Learning Focus: DeepSeek-R1-Zero uses reinforcement learning directly, skipping the costly step of supervised fine-tuning with labeled data. This saves on data annotation and computation.
2. Efficient Training Pipeline: DeepSeek-R1 uses a multi-stage process, including fine-tuning with high-quality reasoning data, reinforcement learning for reasoning, and a second reinforcement learning stage for safety and usefulness. This targeted approach reduces wasted computation.
3. Knowledge Transfer: DeepSeek uses distillation to transfer knowledge from larger models to smaller ones. While models like Llama were used in this step, they weren’t needed for the core training, allowing flexibility and cost savings.
4. High-Quality Data: By focusing on curated, reasoning-focused data, DeepSeek achieves strong performance without needing massive datasets, cutting data collection and processing costs.
5. Flexible Base Models: DeepSeek can use various base models for distillation, potentially reducing costs if cheaper options are available.

Savings:

• Data Costs: Less reliance on massive datasets lowers data acquisition and processing expenses.
• Computation Costs: The efficient pipeline and reinforcement learning reduce computational needs.
• Deployment Costs: Smaller distilled models are cheaper to deploy and maintain.

Based on the sources, DeepSeek could have been trained without using Llama as a base model, although the Llama models were used in the distillation process. Here’s why:

• DeepSeek-R1-Zero’s Pure RL Approach: DeepSeek-R1-Zero is trained using pure reinforcement learning (RL)directly on a base model, without supervised fine-tuning (SFT) as a preliminary step. This demonstrates that the reasoning capabilities of LLMs can be incentivized purely through RL, without needing SFT or any specific base model like Llama.
• DeepSeek-R1’s Training Pipeline: DeepSeek-R1 is trained through a multi-stage pipeline that includes:
  • Cold Start: Fine-tuning a base model with high-quality Chain-of-Thought (CoT) data.
  • Reasoning-Oriented RL: Enhancing the model’s reasoning abilities.
  • Rejection Sampling and SFT: Creating new SFT data through rejection sampling and fine-tuning using this data along with non-reasoning datasets.
  • Second RL Stage: A second RL process aimed at improving helpfulness and harmlessness.
  • Distillation: This step uses the outputs from DeepSeek-R1 to fine-tune smaller models, including models based on the Llama architecture.
• Distillation as a Separate Step: While DeepSeek-R1 used Qwen and Llama models as base models for distillation, the core training of both DeepSeek-R1-Zero and DeepSeek-R1 did not fundamentally rely on the Llama architecture. The key methods for improving reasoning capabilities involved the RL process and the curated training data that was used to train DeepSeek models.
• Emphasis on Reasoning Patterns: The process of using DeepSeek-R1 to distill smaller models is important, as it transfers the reasoning patterns discovered by the larger DeepSeek models to smaller models. This does not imply that the Llama models are critical for the initial training of DeepSeek-R1, only for the smaller models generated in the distillation process.
• Base Models for Distillation: The models used as a base for distillation included Qwen and Llama, however, this highlights that other base models can be used. This demonstrates that the key findings from DeepSeek’s training can be implemented in other architectures.

In summary, the DeepSeek models were designed to be trained through innovative reinforcement learning strategies and a multi-stage approach. While Llama models were used for creating smaller distilled models, they were not a requirement for the core training and methodology of the DeepSeek-R1 and DeepSeek-R1-Zero models. The core of DeepSeek’s improvements are the RL strategies and the curated datasets, not the Llama base models.

Yes, Llama (Large Language Model Meta AI) uses labeled data during its training process, particularly in the supervised fine-tuning (SFT) phase. Here’s how labeled data is involved:

1. Pretraining: Llama is first pretrained on a large corpus of unlabeled text data (e.g., books, websites, etc.) using self-supervised learning. This phase doesn’t require labeled data.
2. Supervised Fine-Tuning (SFT): After pretraining, Llama (or similar models) often undergoes fine-tuning on labeled datasets. These datasets include examples of inputs paired with correct outputs (e.g., question-answer pairs, instructions, or annotated text). This step helps the model align better with specific tasks or improve its performance on reasoning, safety, or helpfulness.
3. Reinforcement Learning (Optional): Some versions of Llama or similar models may also use reinforcement learning with human feedback (RLHF), which involves labeled data in the form of human preferences or rankings to further refine the model.

Why Labeled Data is Used:

• Labeled data helps the model learn specific behaviors, such as following instructions, answering questions, or avoiding harmful outputs.
• It is critical for improving the model’s alignment with human expectations and task-specific performance.

DeepSeek’s Difference:

DeepSeek-R1-Zero, for example, skips the supervised fine-tuning step and uses pure reinforcement learning directly, which reduces the need for labeled data. This is one reason DeepSeek’s training process can be more cost-efficient compared to models like Llama that rely heavily on labeled datasets for fine-tuning.

Labeled data for training large language models like Llama typically comes from a variety of sources, depending on the task and the goals of the model. Here are the main sources of labeled data:

1. Human-Created Datasets

• Crowdsourcing: Companies often hire workers (e.g., through platforms like Amazon Mechanical Turk) to create labeled datasets. For example, workers might write question-answer pairs, annotate text, or rank model outputs.
• Expert Annotations: For specialized tasks (e.g., medical or legal domains), experts may be hired to create high-quality labeled data.
• Open Datasets: Some labeled datasets are publicly available, such as those from academic research or open-source projects (e.g., SQuAD for question answering or CoNLL for named entity recognition).

2. Instruction-Tuning Datasets

• These are datasets designed to teach models how to follow instructions. Examples include:
  • OpenAI’s InstructGPT Dataset: Contains prompts and human-written responses to train models to follow instructions.
  • Self-Instruct: A method where models generate their own instruction-response pairs, which are then filtered and refined by humans.

3. Human Feedback for Reinforcement Learning (RLHF)

• In reinforcement learning with human feedback (RLHF), labeled data comes from humans ranking or scoring model outputs. For example:
  • Humans compare multiple model responses to the same prompt and rank them by quality.
  • This feedback is used to train a reward model, which guides the fine-tuning process.

4. Synthetic Data

• Sometimes, labeled data is generated synthetically:
  • Model-Generated Data: A model creates examples, which are then verified or corrected by humans.
  • Data Augmentation: Existing labeled data is modified or expanded to create new examples.

5. Task-Specific Datasets

• For specific tasks (e.g., summarization, translation, or coding), labeled data is often collected from domain-specific sources:
  • Summarization: News articles paired with summaries.
  • Translation: Parallel text in multiple languages.
  • Coding: Code snippets paired with explanations or corrections.

6. Proprietary or Internal Data

• Companies like OpenAI, Meta, or Google may use proprietary datasets collected from their own products or services (e.g., search queries, customer support interactions, or internal tools).

Challenges with Labeled Data:

• Cost: Creating high-quality labeled data is expensive, as it often requires human effort.
• Bias: Human-created data can introduce biases, which may affect the model’s behavior.
• Scalability: Labeling enough data for large-scale training can be time-consuming and resource-intensive.

DeepSeek’s Approach:

DeepSeek reduces reliance on labeled data by using reinforcement learning directly, without supervised fine-tuning. This allows it to achieve strong performance with less labeled data, making the training process more cost-efficient compared to models like Llama that depend heavily on labeled datasets.

Blade Runner, set in a Tokyo-like place, in 1982.

GM did a bit of work in the 60s with NASA. But I guess they got distracted.

Also – watch Blade Runner 2049. Many issues we will face soon about our AIs. Except next level – the AIs will merge with Bios.

Try it for DeepSeek

https://www.kluster.ai/#pricing

Why is it news THIS week? The stuff was released Dec 26.

E.g., “Agents will be a big deal” vs. OpenAI’s Operator ‘preview’ (which is just *live* to its highest tier customers)

Demos worth seeing: the ‘intended’ use demos, writing and deploying a webapp that would have been a pretty cutting edge web 2.0 startup in 2011,

E.g. “Chip crunch will ease/tighten” vs. Chinese researchers clone OpenAI with 1/10th the chips with DeepSeek (actually it’s a hedge fund’s research team that made it). The API is 97% cheaper than OpenAI, the code is open source.

Demos: cloning Perplexity, this deep dive-ish review by me, and by the way I spent a chunk of time setting up the locally build of Deepseek on my non-Nvidia machine (that failed) in an effort to get BrowserUse to work (will have to try that again later with some different setup for Flash_Attn libraries)

E.g. Crypto’s new Summer is merging with the AI momentum into some kind of Nexus 9

E.g. “MAHA” vs. “What happened to Vivek the Pharma Warrior already?”

•  Latest Models: DeepSeek-V3 & DeepSeek-R1 rival Western AI leaders in performance.
•  Resource Efficiency: Achievements made using less advanced hardware = innovation spotlight!
•  Open Source Win: DeepSeek-R1 employs reinforcement learning without supervised data, making it cost-effective.
•  Geopolitical Angle: Concerns linger about potential alignment with Chinese government narratives.
•  Defying Chip Restrictions: Progress made despite U.S. export restrictions on advanced chips.

 Economic Implications

•  Cost Efficiency: Open-source strategy lowers barriers for startups to experiment, democratizing AI access globally.
•  Market Shift: Could threaten dominance of proprietary AI platforms like OpenAI and Google Bard.
•  Hardware Independence: Success with less advanced hardware reshapes hardware-software investment dynamics.
•  National Security: Raises stakes in U.S.-China tech competition, influencing trade and policy priorities.
•  Global Ripple Effects: Potential to disrupt international AI research funding and alliances.

 Predictions About the Future

1.  AI Talent Pool: Increased collaboration opportunities might emerge between global and Chinese researchers. (Forbes Prediction)
2.  Chip Industry Impact: Advanced AI without cutting-edge chips may shift demand curves in semiconductor markets. (McKinsey Report)
3.  Ethical Debates: Reinforcement learning without supervision could push new governance frameworks. (OECD AI Policy)
4.  Open-Source Boom: Could spark a wave of decentralized AI research hubs. (TechCrunch Insights)

(the above was written by chatgpt)

But check out DeepSeek (a side project of a hedge fund called High Flyer) vs. the rest on this pretty tough task. ChatGPT was better and more relevant (it gave me easy ways to get it done, like a chrome extension I could just click on) but DeepSeek was extremely close, and Anthropic + Perplexity just bonked out.

Fail —

Fail —

DeepSeek: Quite helpful… about the level of ChatGPT last October

Still the best: ChatGPT’s current version. Review it here.

Oh and what was I making? Stay tuned for the next episode of my podcast.

Notwithstanding his mastery of AI… or even interest and expertise in nuclear.

Note the rising rates in US and everywhere. Political clout doesn’t sway bond traders

H/t John Micklethwait of Bloomberg

And the vibes are all about realignment to the new US administration.

Of interest this year for me, and a barometer for entrepreneurial traction against big pharma, big health, big everything…. what will be the suits to sneakers ratio? I remember some stultifying rooms full of suits in past years.

LifeX has two events with 1,000+ RSVPs each… so perhaps there are enough sneakers for startups to conquer this category.

Some cool work. Though edited to be on message. The governing thought of the contemporary section is artists observing how communism and capitalism came to work brilliantly together.

Though surrogacy by a female still necessary.

“A woman needs a man like a fish needs a bicycle.” – Gloria Steinem

I was surprised to hear this view but it’s not fully wrong. Before WW2 – European dominion. After – gone.

But check out the super dope Tokyo they choose as the future. Virtually all of it happened.

Amazingly, GM did virtually no work on flying cars ever. Just a few demos.

Here’s one that’s making me very happy today — Stable Diffusion on OSX — https://diffusionbee.com

You bring your own compute(r), so you can use it as much as you want.

Phil Knight says as much in his book. And weirdly, Tigers never really attempted to tackle the US. Conservative Japanese strategy? Plenty of others were not so meek – Sony, Nintendo, Toyota.

For a start, how do they do it? Everybody has the national health insurance system, and some people get additional private insurance for serious or chronic conditions to get better care. But lots of ‘rich’ people don’t bother with that.

Because everyone has pretty good health care, equitably provided, and because there’s a relatively healthy food lifestyle (seafood and vegetable diet, small quantities, low obesity), it seems people live pretty well and pretty long.

Though the various crises of a very old society are quite amazing, and very much in the cards for the US and EU:

• The politics is inert and nothing ever changes. Most of the voters are old, as are the politicians. For example, Japan is the only post-WW2 country that still doesn’t have autonomy over their own military. Germany, Italy, even Korea and Vietnam are no longer ‘occupied’, but the Osaka to Tokyo flight is still 2x longer to avoid US military airspace. No nationalist has moved to change that.
• Elderly care is a bigger issue than childcare. Adults quit their jobs, relocate cities, and generally restructure their lives to go care 1-1, 24 hours per day for their 80+ or 90+ year old parents. The aged-care system does some things, but not all things. So the economic cost of the elderly population needs to account for lost work contribution of the working-age population.
• There are of course rather fewer young people around.
• Like government, big corporations are basically hereditary conglomerates run by old people. There are virtually zero upstart reshaping the economic landscape. It’s what foreigners often admire about Son-san of Softbank, or the Rakuten or Recruit founders, but there are very few.
• There aren’t anywhere near as many robots around as it seemed there would be a few years ago. Robots aren’t pickup up the empty dishes at restaurants or self-driving cars or doing self-checkout at stores.

1. China can’t afford a war over Taiwan. The country is under too much economic stress.

2. Hong Kong has a long future as the gateway to China. It can’t be replaced by Shenzhen etc.

3. Hong Kong offers a clean landscape for life sciences and food in a way that Mainland cannot. People don’t trust clinical trials from Mainland.

4. The debt crisis on real estate in China is a big deal and will drag the country for many years.

Because of: #resist and protest, country and Americana of a better caliber, high Art, the ingenue Timothee,

The onchain options don’t seems to be trending really. But that would be the likely place.

Certainly not this WordPress blog, right?

Though witness how well podcasts and newsletters are doing.

Maybe it’s just that simple. The return of RSS and linkdumps?

Maybe we have the makings of another pandemic.

Still 100% true.

It often gets mentioned (to me anyway) when founders are comparing notes on how to get moving.
If you are one of those founders who posted about this brief book on some chat, slack, reddit or whatever – please give a shout. I want to update it and you know what it should say next (and be a coauthor!)

Download it here from AmolSarva.com

Alternate: Download it here on Knote

Here’s what the best prediction market (Good Judgment Project at Penn, aka “Superforecasters”) says:

Though the prediction’s forecast has been rising:

The source.

40% is not a toss-up. Clearly the super forecasters think there are some obstacles ahead yet. Maybe in a straight up election head-to-head it would be a toss-up. But add the modest risks of the primary and the legal actions, perhaps that’s the dent in the current 60-40 Biden-Trump forecast.

And here’s another pretty cool one, a tossup:

Here’s that source, Metacalculus.

Easier to check, and sooner, is what happens on Tuesday in New Hampshire. Here’s the prediction – only 18% of forecasters predicting that Trump wins NH by >15%. That’s not a probability by the way, it’s how many predicted what. Chart:

Greetings from the other side — 

It feels like we all need it this year: the other side of New Year’s Eve. 

Many of the folks I speak with were beleaguered in 2023. There were lots of reasons: war, warming, politics. Often for lack of understanding between sides.

One special reason in my business — which is the work of new companies, venture capital, technology and our mission to extend the lives of people and planet — is that it was a really tough year for startups. 

But I think we are getting to the other side. 

You’ve already heard that the stock market ended up doing well, that inflation is OK, and jobs are strong, and that interest rates are going to go down. Public biotechs have been coming back, and bitcoin is up even more. 

Did you also hear that we probably crossed Peak Carbon in 2023, the year of the highest CO2 emissions? Even China, which is rolling out renewables faster than forecast. We might be on the downward side of the slope now. We just need to make it faster.

And medicine: the first gene therapies are now in use, a malaria vaccine exists, new antibiotics, even prosaic stuff like the US Medicare budget seems to have flattened. And GLP1s, eg Ozempic, are about to reconfigure the world while also tackling the top causes of mortality on the planet — cardiovascular stuff like heart attacks and strokes, where weight loss is a tremendous driver.

Of course AI is easily the most significant thing. 

There’s so much opportunity for innovation, investment and company building — I hope if you’re interested or part of it in some way, that you’ll hit me up so we can explore together.

The Prediction Consensus

Meanwhile, I went a bit further on all these predictions.  I read through 50 of the reports that strategists and companies put out, full of predictions for the year ahead. You’ve heard of them (“Peach Fuzz”), but you’d be amazed how many efforts at thought leadership come packaged up for analysis. It’s the investment banks, trend spotters, marketing firms, polling organizations, social media platforms and so forth. 

And here below is my score sheet of the themes that struck me.

// What got highlighted as a major theme in these reports? E.g., a whole chapter heading

• Gen AI xxxxxxxxxxxxxxxxx <– from beauty to recipes to politics to finance, AI was big in nearly every report
• Climate xxxxxxxxxx <– sustainability for products, or just risk
• Mental health xxx <– even in fashion, it’s a theme
• Ozempic xxx <– I think this is going to be so, so huge
• Russia-Ukraine xx
• Israel-Gaza xx <– oddly, the year-end forecasters mostly didn’t touch this one
• Depopulation xx <– population is already shrinking in many countries
• Pessimism x <– scary, the idea that global mood is just negative
• US election x
• AR/VR x <– even with Apple’s new glasses?
• Blockchain x <– “winter”
• Security x
• User-generated x
• Longevity x 
• Doomscrolling zero <– remember the pandemic? TikTok is the opposite
• Fasting zero
• Zoom zero
• Quantum zero <– forever just 5 years away…

// From the banks…they basically all said the same thing

• Economy to improve, interest rates to go down
  • Slow recovery, no recession
  • US strong, China and EU slow

// Public affairs; here are some surprises

• 67% of Indians and 63% of Mexicans expect their future to be better, vs. 48% China and 34% US (& 4% Japan) 
• 75% of the world population (60% in the lowest country) believe we’re headed for an environmental disaster 
• Skepticism about science ~40%  
• Women are far less likely than men to say the Internet is good for society (34 v 47%, UK stat)

// Consumer trends that grabbed me

• Fake/Artificial everything is changing how we see our own identity. As in, “I am actually fake”. 
• More privacy. Perhaps an eternal back-and-forth.
• Consumers feeling unloved by brands. Companies are raising prices and cutting perks.
• Relatable > uber. E.g., normal people as influencers.
• Joy and self-care as an avoidance behavior. Related to the mental health thing.
• Shein, Temu – huge; TikTokShop… Ecommerce is in upheaval.
• Carbon-negative > offsets. Even for stuff like strawberries.

// Tech-related pronouncements that surprised me 

• Bing has gotten way more popular, +35% usage in the UK last year
• Zoom shrank. Google Meet went from nowhere to half of zoom. Slack shrank. Also UK.
• Backstage as the default developer portal for everything. 
• DMs are the new email. I’ve noticed even business people are texting > old-fashioned email (or slack) 
• Millennials will make B2B transact digitally. (More old-fashioned things fading, like PDF proposals from vendors.)

// Food trends that will turn up at the local grocery

• Biocoating, waste reduction. Paper and foldable packaging.
• Brain health as a diet thing. 
• No/low-alcohol stuff.
• Pre-mixed drinks. (The opposite of above.)
• Water is valuable / Drink more water. This will continue to be commercialized.
• Fermentation. Kimchi!
• Protein-everything. Fewer carbs wherever possible will continue…

// Real estate: a few quick themes here

• Interest rates & Construction costs <– it’s all about that
• Office & Retail bad; Housing good <– where to be
• London, Paris, Madrid in Europe – no change
• Sun Belt cities in US – no change

// Health

AI for everything that scientists or doctors or patients do: discover new treatments, diagnose problems faster, get care more easily… This one is huge. It’s a big part of what we’re doing at LifeX, so I won’t go into it here — you’ll hear more from me later. 

All of the above interests me especially in my work, so if you want to get the sourcing on a tidbit or chitchat about one of the trends: I’m interested. Until then…

If you aren’t on my distribution you can sign up here.

AI is moving so fast right now! Well that was a few months back. It felt like folks were discovering whole new amazing things daily. Remember that whole “tell AI to start a business for me and run it” thing?

Suddenly it’s moving super slowly as people crash into the, oh right this “obvious” next step is super hard.

To figure out what’s hard and easy, we had a bunch of really serious health and bio people together in late September in New York at LifeX’s AI x Bio event. Who was there:

• founder of a billion-dollar-plus digital health insurance company, Oscar Health
• founder of billion-dollar-plus AI drug discovery company, Immunai
• founder of yet another unicorn that has one of the largest weight loss apps, Noom
• head of drug discovery for one of the largest Google-spinout quantum companies

plus CxOs of large drug companies, CMOs from large health systems and therapeutics startups, VCs, and multiple more unicorns….

We heard from these folks who are building hardcore and reshaping big product platforms. To me the big takeaways were about what’s very difficult.

1. Short questions. Few followups. While the AI models are trained on complex information (“the whole internet”), they don’t handle lots of input from you very well. They get overloaded keeping too much input “in mind” and start hallucinating. So much of the hard work is long questions — handling an insurance claim involves medical history, multiple doctor visit notes, and complex reimbursement rules. Or a psychotherapy session — or multiple sessions over months and years.
2. Failing inelegantly. The hallucination stuff is when the models say false things with authority. When it happens, the model is failing. But it isn’t caused in obvious ways for the user. One of the ways to make it hallucinate is to overload it. It doesn’t know it’s overloaded.
3. No plans. No nuance. The models don’t think of you as someone to persuade, and in fact don’t remember you at all. Every conversation is the first time. That has its virtues (patience, if you are tutoring someone) but also means “bedside manner” is impossible. It also means the robot will foolishly repeat the same suggestions (perhaps like some annoying humans do).
4. Not enough data. It’s awesome that reading the entire Internet can create credible short conversations, but the entire Internet is a lot of data. Figuring out whale songs would require 40,000x the amount of data we have about whale songs on the planet. That’s likely to be true for conversations between dietitians and weight loss clients, or with psychotherapists, or biologists querying proteins.
5. Inaccessible data. Some huge systems capture zillions of patient conversations, e.g. customer care centers or claims adjustors — but all those vendors and tools and platforms weren’t designed to store every single patient conversation, interaction, and health outcome in a single platform where the robots can train. Big health systems have projects — not to train AIs — but just to digitize millions of hand-written records.
6. Irrelevant data. When two humans talk, the humans react in certain ways. If you train an AI to talk that way, will you get the same result? Unknown. Much of our data can’t be used directly by a robot unless the interlocutors think the speaker is a human too.
7. Disguised data. Hidden factors in conversations — like is it a robot? a child? a foreign language speaker? — influence how the data is usable or not. Should your AI talk the way a teacher talks to a kindergartner? How do you know who the speakers are?
8. Private data. Health is full of private data that nobody is allowed to access or aggregate.
9. Impossible data. The quantum phenomena at the particle level in biology isn’t explained or adequately observed. So some folks are simulating quantum phenomena from bottom-up quantum physics models…and using that data to train their AIs.
10. “Better is best” models. The models got good “suddenly” as far as the world is concerned, and perhaps that’s because there’s a huge visible difference at certain thresholds. But is that at all thresholds? If GPT4 is far better than 3.5, or if the rankings on HuggingFace are crowded yet closely packed…does the winner take all? Most in the room had not even tried Google’s Bard, or any other model than OpenAI in any depth.
11. Memory, computations, speeds. There’s more to life than just the quality. And routing around between models either sometimes or in real time to optimize these dimensions is going to be another decision — some models can take longer input, longer conversation chains, use less compute, react faster…
12. Skeptical users. As amazingly convincing as the models are to some of us, several big players mentioned obstacles in getting patients or doctors or other experts to even trust the AIs at all. “Much easier in the back office where you have fewer people to convince.”

The overwhelming conclusion for me from all this was: choosing precise applications that actually work/impress/save time + generating your own private data from these applications is going to be the path to greatness in the near-term

Here’s a recap speaker-by-speaker

Also: our next session like this, AI x Longevity at Betaworks on October 12.

Get on my newsletter here for future posts

Here’s an awesome example.

I was walking in Brooklyn once, some years ago (don’t know when), and I looked up an address online that seemed super weird. The building was scary! Online it said the building was being prepared for a film shoot with Stephen King.

I tried to remember today what that moment was.

Screenshots tell the tale.

Bing’s ChatGPT 4 implementation was so essential to finding this thing… Yes Google still had a few clutch moments.

There is a way! And it doesn’t require buying an app.

It’s because all your attached images/movies live locally on your computer in an accessible place. There is a folder in your user’s home directory called Library, in that is Messages, and in that is Attachments. And in that are 1000s of folders for all your messages with the various attachments inside. It’s not really practical to gather up your pictures from there using the OSX Finder interface.

But you can type a few lines into terminal and copy them all out to a folder on your desktop.

What you need: a Mac that’s linked to the same iCloud

How to:

1. Open Messages on the Mac and make sure it’s downloaded all the conversations etc
2. Make a folder on your Desktop called MyMessagesAttachments
3. Go to System Preferences and into the Security app in there, and give Full Disk Access to Terminal. Now Terminal can let you do things in folders that Apple considers unsafe
4. Open Terminal
5. Type “sudo -s” (it might want your password). You are giving yourself permission to be a super user. Be brave
6. Type “pwd”. It will show something like “/Users/username” <– “username” is your username on this Mac. You’re going to replace that word in the instructions below with your actual username
7. Now type this: find /Users/username/Library/Messages/Attachments -type f -size +100k -exec cp {} ~/Desktop/MyMessagesAttachments/ \;
8. It kind of runs for a bit then there is a folder on your desktop with all your pictures and videos (all files greater than 100kb in size actually, but now you can manage it all way easier)
9. Now type: chmod 777 /Users/username/Desktop/AmolsMessages/*
10. Go to your Desktop in Finder on your Mac (in the normal UI) and check out that folder…
11. You an change the Messages settings now to remove old stuff (it will remove old texts and images and videos…so you are still losing old chats, sorry)

You don’t get any metadata with these images; they are all stamped as created just now.

But that’s good enough for me…. Enjoy! (Hit me up by the way if you have any issues.)

Also, I suspect you can just ask Bing/ChatGPT to walk you through this. I asked it lots of ways for help with the *problem*, but at some point it occurred to me that using the file system was a good idea and I typed this in below. That led me to the answer.

But since covid they have been failing to compete – just way less liquidity and far higher prices in the US. Here is the NYT takedown.

And it’s almost 6 months later today, 308.

Here’s the way you can value the stock of Apple, based on the present values of the cash flows you think Apple will generate (discounted cash flow):

Apple earned about 100bn last year, so you just need to estimate how that annual number will grow over time, and how you want to discount future earnings. The snapshot above shows 4% growth, and a discount rate of 9.8%. The discount rate being the key here. You are assuming any dollar you *don’t* put in Apple stock could earn 9.8% somewhere else, without any risk at all, and so a dollar that Apple promises you next year is only worth about $.92 today (since you could earn 9.8% on that $0.91 elsewhere and it would be $1.00 next year).

That discount rate goes up when interest rates go up, and though people mostly talk about the Fed rate or mortgage rates, there are other layers of lending products out there that also move in tandem with interest rates. Check out the Bank Prime Loan Rate in the US, which is 7.5% today. Corporate bonds in the US are actually around that too, and Emerging Market Corporate Bonds have been 10-12% lately, because they are more risky I guess. (More trends.)

The risk-free rate of return is actually a straight-up ingredient in calculating a discount rate (aka weighted average cost of capital), so higher rates make discount rate higher. And another one is stock volatility (aka beta), and lately the fluctuations in the stock markets and startup markets make that higher too.

In normal times, you’d say a public company like Apple had a 10% cost of capital and maybe startups would be 20% or higher.

Venture lenders were charging startups 8%+ when corporate bonds were 4% back in 2018; and lately, the venture lenders are charging 12%+ when corporate bonds are at 8%. So that’s kind of a ballpark ratio. (Cost of capital includes the cost of debt but also the cost of equity, which I won’t get into.)

So what’s a company worth?

If you mess with the discount rate only, in a calculator like this one, and with numbers for Apple, you can make Apple’s valuation swing wildly from 1 trillion to 2 trillion (where it is today) to 3 trillion (where it was nearing recently). Just move the discount rate from 7% to 11% to 16%.

Take a look at this sheet I made and play with the numbers on it, except this time for startups. A massively successful startup that makes cash flow earnings in just a couple of years, and grows phenomenally for years and years…in a high discount rate world, is worth not so much. That’s pretty much what happened to all these high-growth recent IPOs – they don’t show earnings for a few years (because of the growth), but in the more distant years where they do have earnings, the bite of a high discount rate devalues them.

If you play with your growth stage startups forecasts in that sheet, the discount rate can swing you from 200mm valuation to 2bn if you tinker with 7, 11, 16, 20%. (If you are a public company, don’t you think you’re closer to 11% than 24%, after all? Perhaps not.)

When people talk about a “taste” or a “sentiment” for stable earnings or traditional business models, I think they are foolishly ignoring the plain old math of returns. When rates are a higher, earlier income is better. Which is why traditional industries have fared better this last 18 months (the Dow isn’t down nearly as much as the Nasdaq.)

Startups valuations

Which comes back to the startup valuations – at pre-seed, seed, series A – that founders care about. Fred Wilson put his view out yesterday that we should see valuations return to 2015 levels, where a seed round is around 10mm and an A round is around 20mm pre-money. That would be about half of where things were 18 months ago. Nasdaq finished 2015 around 5,000, and is around 10,000 today, but 15,000 a year ago. So Fred implies that the tech stock market is oversold.

On the other hand, Howard Marks seems to be saying the opposite, as he’s usually a bear.

If we think the tech stock market will come back up in the coming year or two, the main assumption driving it would have to be that the discount rate eases up. Having browsed lots of year-end bank forecasts, they mostly agree that inflation should stop next year, rate hikes should stop, and even if rates don’t get lowered, it means markets will start showing lower rates on bonds.

Lower rates, lower conflict.

About predictions

The funny things about predictions, though, is that nobody predicted Russia-Ukraine last Jan 1st, nor did anybody predict China would spend the whole year in oblivion vacillating on Covid. So I suppose the real value of predictions getting some insight into a theory of what makes the world work, and using that model to understand new things as they happen. Here’s hoping that prediction is true.

Hear the corporate raider playbook – fire the “nice guy” CEO, slash headcount massively as most of it is asleep, and promote “smart” people who challenge the groupthink internally.

It’s interesting to hear from Icahn how much smarter he thinks he is than everyone else. You kind of have to take his word for it, since the talk he gives is pretty light.

I suppose he is smart, but he sounds like a an Ivy League-elitist of the Vanderbilt and Carnegie ilk (he went to Princeton and dropped out) more than a table-thumping barbarian (he did the RJR Nabisco takeover). Listening to him (it’s 2009) you’ll hear the words “make America great again” in a jumbled paraphrase, and quite a bit of Trump-like rhetoric.

In the Swensen “Yale system”, the Buffet idea of “just own the index”, seems not at all what made Yale succeed.

Here is a summary of his talk while visiting Schiller’s investment class at Yale in 2010 or so.

First, a link to his basic talk on strategy and how they succeeded: https://podcasts.apple.com/us/podcast/financial-markets-audio/id341651121?i=1000063752200

An interesting ranking he shares of returns by class:

VC +45% per year, 1st quartile vs 3rd quartile

RE 17%

HF 11

Equities 2-4

Bonds 0.5%

His list of priorities:

Equity first and diversified

Don’t time the market

Security selection – choose your skills-match and consider the efficiency of markets you target

Yale therefore is 70% real assets and illiquid equities, plus 26% in public equities— 20%+ in private equity and venture. (Compare that to most “normal” people or family offices…)

Results

⁃ topline has been up a lot, 22x in 25 years

⁃ with only 1987 as a down year (-1%) and Yale was positive every other year

⁃ equity drove the returns but diversification made it stable

In fact, even this August, the excellent Elizabeth Colbert of The New Yorker appeared on a program talking about George Bush saying back in the 1990s that Americans would not compromise on their way of life.

A few years after everyone wore masks, closed schools, stopped commuting, gave up international vacations…not just temporarily mind you. The talk of today is still the end of the office (it’s been where things happen since the 1920s), the end of “downtown” (it was the definition of cool for 50 years), face to face meetings (“the room where it happens” as it is said in Hamilton), and more.

Behavior change, in short, does happen. Not just that: we take car shares, look at smartphones, ebooks, video calls, avoid sugar, count steps, etc. Behavior change happens all the time – we call them trends or fads.

In a moment where the US climate change policy is in the spotlight, the amazing thing about these commentators focusing on behavior change, is that the climate policy that seems most likely to work involves very little scrimping or saving.

We’re just going to make it cheaper and easier to emit zero carbon.

Technologies ABC – green electrons, electric cars, solar panels for heating and cooling = no change to the suburbs. Synthetic meat, new livestock feeds = no change to carnivorous life.

I’m optimistic

Why did we think this? Well, most obviously, because the market has been up for the last month after terror since Nov 2021.

The terror was…fall of meme stocks, fall of covid darlings, signs of persistent inflation, higher interest rates and signs of more high interest rates over time but slowly rolling out (e.g. Europe is raising at a lag, not decisively), which means lower liquidity, Ukraine invasion, risk of breakout, risk of major war, war contagion to China eyeing Taiwan, China lockdowns, Covid part 4, monkeypox, chip shortages for cars and appliances and everything, US government paralysis, our own political thin skins hurt by Supreme Court overreach and Joe Manchin, climate crisis heat wave across the map, skyrocketing oil prices, risk of massive recession and tech stocks in particular getting crushed, causing venture and startups to meltdown, and therefore a general buzzkill.

It was really bad out there in July!

But we agreed on a few things, kind of factual things, that might be behind the feeling that the worst is over:

• Russia is stuck and won’t be invading anyone *besides* Ukraine
• Ukraine is at least at stalemate, and is acting on offense
• Oil prices have started falling, weakening Russia
• China is chastised; won’t want to try that gameplan in Taiwan
• China finally flinched on lockdowns and is opening back up
• Inflation ticked lower, suggesting a top, end of rate hikes, return of liquidity…at some point soon, not indefinitely in the future
• Which also means we weren’t “just getting started” on inflation, it maybe was transitory after all (though over 1 year, not 3 months)
• Covid continues to ease
• Monkeypox isn’t too bad
• US politics: voters struck back on abortion, and polls show big momentum for Dems, plus Congress/Biden passed a climate bill + infrastructure bill + health care bill, and are generally doing stuff
• Companies are doing lots of business, hiring shortages prevail (high employment, wage growth), shortages of lots of stuff (gas, cars, chips), and this all means the economy is very strong — so no recession signs at all
• Which all means: maybe this latest economic crisis was yet another ‘flash crash’ like March 2020 and Sep 2019 and Sep 2018 and all that.

Maybe by January we’ll be back to the January 2022 mood…

One of the recent philosophy breakthroughs, IMHO, has been analyzing the possibilities of artificial intelligence (check out Nick Bostrom).

Another one I came across recently was “ethics across generations”, which has been around a long time but has a trendy new name that might get some traction: longtermism (it sounds like a stock-picking philosophy.)

What should you care about, really big picture, if you care about the future of humanity?

An interesting point of view from my friend who works for a space company. They are making a space station (kind of like SpaceX replacing the launch rockets) as the current one is coming down in the next 5 years.

Apparently, space people fall into camps relating to this question; it’s the reason they work on space. Here are some of the camps:

– we need to be interplanetary for humanity to survive, b/c Earth may fail (asteroid, climate change) – e.g. this is Elon with SpaceX

– we need to develop space tools so we can move our dangerous activities (energy production, materials synthesis, pollution, work away from our home (Earth = sanctuary)

– we need space to learn faster the solutions for ourselves (science in microgravity could reveal big breakthroughs that help us live or fix planetary issues) or we can get resources there (energy, minerals) or perspective (e.g. the “whole earth” campaign of Stuart Brand)

– these same space people are also often Earth people – e.g. Elon with Tesla

– and human longevity people – e.g. Elon with Neuralink (where a brain-computer interface is the way to create a non-biological path forward)

To me the interesting insight is: if you care about humanity, you need to care about Earth, and if you care about Earth then you might find yourself also caring about space. But it starts with humans, not just science fiction.

Nick Thompson’s stuff to read. He’s CEO of the Atlantic and generally reads a lot of weird, interesting journalistic stuff (plus stuff about running, like a piece about Malcolm Gladwell doing a podcast with a running publication.)

The Goldman Sachs equity markets “On our minds this week…” by Dees and Esposito. It captures the Wall St chatter pretty well. (This week it’s “there might be a soft landing, but not sure yet.”)

The social-impact-minded lawyer Kyle Westway’s Weekend Read. Pretty charming, calm visual style and brief content writeups.

Our new company Aikito has a great one on proptech – short, insightful, and sings the praises of using location financing to smooth your capex budget for new buildouts. Here’s the newsletter.

My own personal newsletter is a very occasional “what’s new in my world” update – it can cover my new products or my new venture investments or Davos or great books. Here you go.

My longest duration relationship with a newsletter is Benedict Evans. I used to read him and Asymco and Daring Fireball. My browsing changed but my newsletter consumption didn’t.

Oh and my occasional newsletter that you can get on by emailing me or signing up here.

To do: someone needs to explain to me how a substack is different than a blog-that-gets-mailed-to-you. What someone should really be doing is re-inventing RSS….

2019 is long enough ago that this paragraph from the book feels prophetic, look for “super-spike”:

Maybe that was Putin’s logic for the recent invasion of Ukraine; his hand is kind of strong now as folks are reducing fossil fuel investment rapidly, but it will keep getting weaker so best to hurry.

It’s really worth reading but here are a few shortcuts I thought were relevant for proptech types (like me).

There’s a detailed section on carbon from the buildings industry and his vision of the future for it. Lots for us to modernize.
1. Buildings are the #1 source of carbon (residential and commercial heating mostly, and increasingly cooling). #2 is cows/beef and #3 is transportation/cars/oil (cars are 70% of oil use)
2. That needs to switch from gas/oil building heat to “green electrons” (use electricity, and make the electricity from from renewable) –> so buildings need to replace gas fired boilers/heating (and certainly oil, which I think lots of old home still have)
3. The buildings also need to start producing energy: solar panels everywhere, storing power: batteries, and sending power back to the grid: make more power than you use. This decentralized power generation is very valuable because energy transmission over long distances loses 70%+ along the way; so if you make it locally it’s 3x as valuable.
4. The buildings need to be smart, have sensors and controls, so they change their usage based on what’s going on, so we don’t hit peak loads as easily — since the peaks really drive the whole system more than the ‘averages’.
5. Buildings of course need to NEED less energy also – upgrade insulation, windows, leaks, waste everywhere in the system. That’s it!

The nice thing about Rifkin’s writing is how totalizing it is, a complete unified theory that builds on itself over time. Everything shows its face again here on Green New Deal: “Beyond Beef” from 1990…”The Zero Marginal Cost Society” from 2015, “The Third Industrial Revolution” (2011 – and no there is no fourth one yet, Klaus, as he reprimands the WEF here in a very digestible 1 hour talk on these topics.)

As for cars (electric + ridesharing + autonomous = way fewer cars, but more utility/use for everyone and 100% less oil), and for cows (1 unit of vegetarian protein uses 1/20th the energy as beef, so basically switch to vegetarian meats and we won’t have any more cows to deal with), the answers are equally straightforward to understand and very “technology and innovation will solve it”.

Oh and here’s something amusing — definitely better to wash your clothes cold all the time (90% less energy than hot water) and apparently dishwashers use less water than hand washing once you get to 8 dishes in the dishwasher. But those two facts have been true for years and the behavior change is still pending.

The other one is top-down, and it’s quite amazing how much wisdom Graham casually tosses in alongside his method, without ever really addressing the other way (he calls it “riding the wave” vs. “being the wave”, kinda.)

His view is “know something interesting” because you are interesting, or at least astutely figure out what makes you interesting. You could become interesting of course – like study genetics or computer science or hang out at the cutting edge is happening, get into Stanford, be the coffee runner at Tesla, whatever. Or you could notice that you are unique (and therefore interesting) because you are young or living through Covid or living some weird place far from the conventional wisdom of New York or Palo Alto.

This view is pretty similar to Peter Thiel’s about having a contrarian and distinctive view rooted in your identity and experience.

It’s where Aikito comes from — location finance platform for companies leasing commercial real estate. I spent five years running Knotel, spending up-front capital to make soon-to-be-profitable locations, often with customers in-hand. Why wasn’t there someone who would finance those revenue streams linked to physical locations? There are lots of peculiarities of real estate-linked costs and revenue that make it hard for banks to do. But there are lots of companies like Knotel out there — they run medical offices, retail chains, restaurants and food courts, hotels and short-term rentals, flexible offices and coworking, dark hubs for logistics, renewable energy sites, and more. So: Aikito. “Make what you know.”

The other school, which I think is often (in my circle) associated with Oli Samwer and Rocket Internet, but also with Fabrice Grinda and Jose Marin and others, is an equally interesting one: when you find a business with a model that works, you can vary it’s configurations to find more businesses. Change countries, change the customer segment, change the price tier, change the particular product itself, and so forth.

For my work this last year, CornerUP is an example that’s much closer to this. Many US players have been building instant commerce facing consumers, but some developing world players have been building business-to-business platforms, so small business owners can source and stock their merchandise more easily. In fact, there was only very impressive player in the US kind of like this — Faire — and virtually nothing else in the US. So we thought: let’s zoom in on the very local b2b marketplace area. The model makes sense, the US is different but has similar-enough dynamics, let’s figure out how to make it work.

Some start top-down, some bottom-up.

To make Aikito work, though, we had to go zoom out and learn how the top-down fintech world operates, and as we figured it out, now Aikito feels like a “b2b fintech for proptech players”. Something you could dream up by crossing PIPE/Capchase/receivables-financing/BNPL with Knotel/WeWork/Industrious.

And to make CornerUP we had to go learn the corner store business, talk to folks, hang out with them, sell to them, hire them in, eat lots of chips, etc.

The duality of this approach is very much behind Life Extension Ventures, which starts with the top-down observation — noticing “a wave” — and is now starting to deal with individual companies, founders, and ideas as we figure out which of their models could work on narrow problems and can be generalized into big, huge companies.

But the vanguard is off doing the human + machine thing with stickers, needles and RFID – which is to say I am finally trying a continuous glucose monitor (CGM).

You put a pretty small sticker (it has a super thin needle like a mosquito’s stinger) on your arm and it tells your phone every five minutes what blood sugar it sees. I feel like I am the last one to try this, as lots of tech folks I know are also secret biohackers.

So what blood sugar changes are interesting to know about?

“Healthy” is low and stable blood sugar of 70-100 mg per dcl.

The world we live in, in the US, is most people are at risk of diabetes – so “healthy” is nonsense.

So…what have I learned in 10 days of watching this thing?

Exercise (running, pull ups – real exercise) raises your blood sugar. because your liver makes energy for you I guess. Eg 100 becomes 180.

Nightime spikes. Not sure yet why but at 2am my numbers go up. 150 or 170 and only very slowly go down over 5 hours.

Morning spikes. And then they go up again! To help me have energy for the day? they stay up till noon.

This is all when I am fasting. I did a 2 day fast – and still the glucose fluctuates like mad.

Chilled out exercise – like walking for 5 minutes, doesn’t spike. It just uses the glucose in my blood and slowly lands the plane.

Fasting doesn’t add to my unpredictability, but when I eat at the end of it, the numbers go wild. What’s the point of fasting if there is no feasting?

My next experiments are going to be a) frequent light exercise, b) way more fiber to reduce the impact of carbs, c) less feasting, d) more regular intense exercise (run daily). And I’m going to try to get some CGMs from India since the US ones are overpriced.

Stewart Brand is how I learned about that – because Brand’s early life was all about domes! It was a mid-20thC sensation.

Like drones for us. Many founders went that way in 2015. One that I backed called me yesterday to report a vicious down round. Drones were a flub for the startup community (less so for the US military).

AI…for startups has produced fewer winners. For such an important technology, there is no FAANG. Except the incumbent FANNMGs who did benefit.

Final note: Facebook is the only one of the acronym where nobody writes books about leadership and management inspiration. The rest are covered by titles touching Bill Campbell (Apple and Google), Bill Gates, Reed Hastings and Freedom and Responsibility, and six-page memos lore.

Running this AM, after a few weeks of hiatus, I realized something. Well a few things — which is the point.

First of all, I had a blinding insight early in the morning. I’m lately shaking off cobwebs and hurrying to comb my hair before my first video call. Not today.

In fact, I had three blog posts I wanted to write before I got my shoes off.

The blinding insight though was what I had done yesterday — sit on video calls all day at home. At the end of the day I was remarking, “feels like I was on a long flight”. A year ago at this time people would comment on my lovely home office, and how I was pacing around on it constantly. Standing, moving, shifting positions. Yesterday and for the last number of days I was a static, note-taking, long-haul passenger.

I was changing into a sedentary person in just a couple of months.

No commute, no morning in-person meeting, no reason to get up more than 15 minutes before it starts, no movement or activity between dawn and the start of action… and no people, faces, fresh air or sun or sweat.

Work-from-home and work-from-anywhere, both, leave it up to you to live *in the world*. For a lot of Covid, I was doing it well. The last few months I saw how easily you can slip into the trance.

I’m going to shake it off.

It’s about how to use your internet/tech chops to start a company that can actually move the needle on climate change…without you necessarily having to go learn how to build fusion reactors. Founders who want to tackle climate change have a lot of vectors they can use. Go for it!

Can we live forever? Perhaps an odd question to be asking in a time of Covid, climate crisis, and war. 

But since you know me, you won’t be overly surprised that I think the answer is yes, and that I want to be part of making it a reality.

Over the last year, I’ve been working with amazing collaborators on new companies that transform the climate impact of buildings, the local food supply chain, and how the next generation joins this work. Popular change, we hope, as these become large businesses.

Now I want to share with you the next step in this: Life Extension Ventures, a fund to back founders who are building companies that put tech and science together to extend the life of people and planet. Longevity and climate. 

Investing in Life Extension

My friend Inaki Berenguer approached me a year ago to say: we’re both PhD academics turned serial founders, who have raised $1bn+ of capital with multiple exits, and have also invested to support 100+ founders…shouldn’t we work together?

In fact we have to do it. We owe it to pay back the good fortune of being trained at some of the world’s best institutions.  

We are working on places where tech accelerates innovation, and where tech accelerates adoption. Science is finding big answers, and tech — software, data, models, and networks — can accelerate it. 

1. War in Ukraine
2. Climate
3. Psychedelics
4. Quantum computing

I would not have predicted 3 or 4 either, but I did dedicate a lot of last year to climate, and an ever increasing amount moving forward as we prepare to launch a new fund dedicated to life extension.

But the most interesting prediction I’ve heard *since* Davos is from my Dad: that not only might Russia lose, and Putin be toppled (by 12/31/22), but *also* that Russia might break up…into its ancestral ur-lands.

What such lands? Well, one is Circassia. It existed when Lincoln was president. Read the link – the languages and place names are positively Tolkien.

I am chatting to a founder who just sold his company. I need some ideas, he says. How should I go about this period? What have you, Amol, been up to?

Well, this is my gameplan for how to create the next big thing.

Re-energize. Running Knotel was pretty exhausting. Large company, lots of timezones, bureaucracy and departments and politics and stuff. Some of my colleagues went to re-energize by simply taking time off. That doesn’t help me get energy though. It makes me fidgety. So for me it was getting right back to the drawing board on new companies. Figure out an idea and a market space and start figuring it out.

Connect.

Go top down.

Read books.

Use what you know. Content vs. Frameworks.

If the content is the same, and you think you are an expert on it, challenge yourself to think of a 10x bigger idea.

What’s very likely to be totally different by 2030?

Gender identity and gender roles, at least in US and EU.

Climate crisis. We will have a solution underway, even as fires burn.

What’s probable to be very different by 2030?

Age of American supremacy over. Great Powers realpolitik returns.

Capitalism, political liberalism, democracies as ideologically sainted values over.

Socialism not back.

Great AI and labor forces panic over. Labor shortages.

Big economic redistribution programs underway – a version of reform without revolution on income inequality, possibly through new services from the state (education, health care, disability, video games and entertainment, cheap goods/food).

Russia cracks, 1989-style.

China stagnates, Japan-style.

But the start of the year included my friend from college N, who died last week.

And reflections with friends who lost multiple relatives in 2020 and 2021 to COVID.

And the passing of time that steadily frays all healthy people.

Best wishes for a better year ahead everyone. My optimism is always there and always about the future we face.

Highlights

The playlist of greatest 5ths on Apple

Part of my process is reading books and I recently finished two that shed light on what later stage startups need to think about — being public.
Most founders know little about it, get a BS class on it from self-interested parties as the time arrives, and some founders get robbed (by debtholders…in my case) when that flank is not guarded.
There a loads of investing books and the best ones overlap a ton – anecdotes that flesh out an approach to evaluating companies and being contrarian enough.
Hedgehogging by Barton Biggs has some things to offer about markets — asset allocators / LPs, ups and downs, burn out, tips, and the types one meets on Wall St.
More generally useful is One Up on Wall St by Peter Lynch, which many say is just the #1 best Wall St book. It might be. I just finished it and already want to read it again. It is meant as a stock pickers guide, but it is effectively a course in how to express and execute strategy so it makes sense to public markets. You won’t find startup ideas here. But you will find the checklist you must meet when you are big – the natural laws you will want to play nicely with. That checklist is like 100 things. At the link above.

Lasch – culture of narcissism

7 new photos added to shared album

Cultural classic from around 1979. Nietzschean. Explains a strong aspect of white power conservatism of this last few decades.

The main point is: Americans have become wimpy and they hate themselves.

Why? It is hard to succeed in America as of 1979. No frontier, no payoff to the Protestant work ethic, diminished individualism in face of corporations, factories, bureaucracies and jobs. And frustration in political movements – no greater cause to be part of. The 60s failed, the little guy stays little.

So people turned inward: yoga, self-help, therapy, meditation, jogging. They also lost belief in the future (won’t be better, doesn’t matter – nukes, climate, oil prices, globalization.) And lost belief in the past (doesn’t help me, no legacy, the rules changed.)

Inside they found insecurity, and look for approval, positive feeling, validation. They didn’t have a fundamental source of personal pride and worth that religion once gave, nor a clear moral code to live under, or personal success to chase in an age of diminishing expectations.

This is what Lasch means by narcissism. You want to be loved by others. The book was reprinted under Trump – a guy who seemed to stoke these fires.

What to do? Well it doesn’t have a formula. Neither did Hillbilly Elegy. But Lasch’s other great book, about the myth of progress, makes the worthy point that the “progressive” value system does lack a few things that some people need. We don’t have a clear, unchanging value system passed down through the generations — or so it seems when our views of what is OK or not OK get updated a lot. There isn’t a legacy from our ancestors when we tear down the old monuments or redraw old class categories.

We should be updating our views – those monuments are to bad things, mostly. But I think it is also true that they meant something to some folks, and those folks didn’t have much else to hold on to as values and self-worth.

Next up, another conservative core curriculum text: Jordan what’s his name.

What have I been doing? I’ve been talking to lots of folks and reading things and trying services/demos.

I’m kind of pleased how quickly a handful of really big, promising, doable concepts are testing out for me. More on that at a high level here at the knote on Popular Change (the umbrella for these efforts).

The market selection, the mode of attack, the ‘founder-market-fit’ and the positive vibes from the venture community are all terrific. Plus I’ve started actually doing stuff on each and they are getting market validation from real market participants. One of them is at the point where we have the first few customers and a flywheel design that is one of the coolest I’ve worked up.

So I’m focusing a lot on people the last few weeks — and perhaps the next few. Having become an expert on the first five levels of the startup video game, and having been to level ten a few times, I think this is a good time to very intentionally and patiently mine/develop my people circles.

If you know folks who would be interested – send them my way. And especially if you know folks from some of the traditionally underrepresented communities in founder world — female, Black, latinx, and so forth. More on that here.

Ship While You Sleep

No Description

Learn about Code As Cards, and everything as cards.

But not when you looked at important further aspects like compensation or rank. Go higher in rank or comp, and see less diversity.

One of the reasons this happened is because the highest ranking (like me the CEO, or the top 2-3 execs) were also the longest serving. I started the company — and I come from a group that is pretty advantaged in elite corporate settings.
And the first few people who joined Knotel? Folks I new through personal circles, and also mostly-advantaged by background.

When you get focused on fixing diversity at the top ranks of your company, but the top 8 execs are all male, all privileged backgrounds…well, it’s a tough and long road to fixing it.

We did a whole series of important diversity leaders from Netflix to Pfizer to Asana last year, and the headline lesson was set bold targets. (Have a look)

So here I am back at the first moment, starting a new company. Should I leave diversity of the company for another day? After we get going and we know we have found a core product/market fit? When we can focus on non-existential questions?

Or should I make a diverse team part of how we conceive of and build an important new contribution to the world? Startups are a chance to make valuable change — shouldn’t we get a lot of voices there right at the outset?

So if you are wanting to be part of a new company, have been thinking of starting one, and if you have been trying to start one — especially for those of you who are different from me — give me a shout. I would love to hear from aspiring entrepreneurs who are Black, women, latino/latina, LGBTQ and others too. Door open. Hit me up by email to a at sarva dot co

More context at my linkedin post as well

RSVP here before 11am ET Tuesday — https://knote.com/p/DBdFOw

I am among the endless entrepreneurs who have dreamt over the years of a personal CRM. Well it exists. It’s HubSpot.

Hat tip to David Teten.

Hubspot lets you:

• Have lots of contacts
• That sync activity against the names
• In an inbox you can assign to a colleague
• Plus scripts
• And apps: forms, lists, chat, phone calls

Eager to dive further in. It’s what I wish LinkedIn would be.

Process

The 2021 Ideas List

Companies

Aikito.co

CornerUP.com

Ratio21.com

Dinara.com

Frank Slootman of Snowflake

Anil Dash on inventing social media and remembering social responsibility

Saeju Jeong on creating the Noom app for healthier lives

Abdur Chowdhury on guiding Twitter and post-social media

Jared Hecht on creating GroupMe and re-inventing small business lending with Fundera

Some really amazing guests and heroes I got to interview this last 12 months:

Legendary Harvard Kennedy professor Ronald Heifetz on leadership

Jeff Sutherland, the inventor of the Agile methodology

Legendary biologist and CRISPR co-inventor George Church

Geoffrey Moore, innovation guru and author of Crossing the Chasm and Inside the Tornado

Superforecasting author, Phil Tetlock

Martha Stewart

Super producer Money Mark

Co-creator of the TED videos, Ronda Carnegie

Stephen Wolfram, computer science god++

Oil prices are going down.

They are today lower than the bottom of the Great Recession fall. Partly because the US has continued to produce more, and probably the relative growth of some renewable use. Oil use globally growth pretty slowly — it’s been around 100mm barrels per day for years, up from 95mm to 103mm over the last 12 years.

2008, the price of oil fell. What happened in 2009 to venture capital investments?

It went down in 2009. Notice it went down in 2016 too — the year after the oil price crash in 2015.

VC and oil exist in a large market context with many economic factors — stock market, real economy, jobs, currency.

But they are directly linked too, something entrepreneurs and VCs don’t often talk about (if they think about it.)

Oil sales (100mm/day x 365 x $50 = around $2T/year) –> Middle Eastern sovereign wealth funds (Qatar, Kuwait, UAW, Saudi sovereigns are ~ $4 trillion) –> investments in funds –> investments in public and private companies.

The NYSE and Nasdaq combined are around $40 trillion of market cap. So these 4 oil countries matter a lot. They are ~5% of the money in the US stock market I would guess. But not just stocks.

Big primary capital groups have some mix between real estate, public stocks, debt and private equity. Yale has 19% in private equity/venture capital. They have a reputation for being pretty heavy on illiquid assets like this (7% in real estate, 7% in natural resources), so maybe the average of most smart primary capital pools is half, around 10% in venture/private equity.

Those are big US university endowments, but it’s a way to think about how the $4T of oil sovereign wealth might be managed. They have $400bn in venture capital firms + PE (let’s say 30% in VC and 70% in PE). That’s $125bn invested in global venture funds as an LP — invested in funds that invest over say 3 years, and harvest over 7.

Venture Capital is around $250 billion per year recently for the entire world. Let’s say the ‘market cap’ of all VC-backed companies is 3x that in the private markets: $750 billion. Around 1.5% of the public markets. So it’s a lot smaller. We just estimated that oil money is 5% of public markets.

What is oil money as a share of venture capital? Well funds invest over ~3 years and hold for returns over 7 years. So a brand new ABC Ventures Fund 8 raised in 2019 will have spent all the money in 2022, and will need it again. About 2/3 of all funds will be on the market in a given year trying to raise a new fund this year/next year.

Since $250bn is invested per year, or $750bn to be invested by the current cohort of funds — what’s the oil share? $125/750. About 15%. Three times higher than the public market. And it means these groups anchor new venture funds — they don’t round off the venture funds, they are the biggest blocks around which many funds assemble the rest of their smaller LPs. A $300mm fund will have a few $30mm checks (sovereigns, pensions, etc.), and a bunch of $10mm checks.

Which means…when oil falls from $80 to $40 as it has in the last year, the job of raising a venture fund becomes super super hard.

And next year, you the startup founder will notice it. Even if the crisis looks over in the newspaper. Whiplash.

Oil groups are sensitive to the price of oil. A unique variable. But ALL funds — pensions and sovereigns and sovereigns — are sensitive to the price of public equities. Most are ~50% in the stock market. When the stock market falls 20% as it just did, folks have bigger fish to fry than the far smaller positions they have with longer exit timelines.

So the advice this week from Sequoia — RIP Good Times II – that coronavirus is a huge risk to startups — is informed by advice to venture capitals: VCs are going to walk through the Valley of Death the next 18 months.

One is Bari Williams – a Silicon Valley all-star lawyer who happens to be one of the few African-American women in one of the ecosystem’s that struggles most with diversity. We get into it.

Stacy Schiff wrote one of the best books I read last year — Cleopatra — and led me to a few of her other fantastic titles: the biography of Nabokov’s wife, and the story of the Salem witch trials. Cleopatra was a fantastic military leader, diplomat, queen, the richest person in the Ancient world, linguist, and more. Vera Nabokov was an essential partner to the great writer. And the tale of the witches has much more going on than the male historians have managed to tell so far. We talk about how significant contributions by women in history have been remembered.

Seed deck:
Need just design?  Use Konsus.  Cheap, fast.  No idea partner, just execution (spend ~500)

Need design and familiarity specifically with seed decks?  use sketchdeck.  They’ve got templates and can force you a bit to think through it following existing conventions.  (spend $1k-2k)

Want freelancer?  Use upwork to find one.  Follow Amol visible work and pay for ~2 reasonably representative slides from 2 different people at once (all-in spend $150).  Work with whomever does it the best.

This is the best guide to seed decks I know of: https://blog.ycombinator.com/intro-to-the-yc-seed-deck/

And while I’m throwing around YC links, my favorite hack to tease out the content of a seed deck without spinning your wheels on design decisions.   Just apply to YC.  The application is the perfect way to force you to think and verbalize your search for product-market fit.  Waaaaaaay better than what most people incorrectly think a seed deck should be, which is a ~2003 version of a business plan.  Cost is zero.  Iterate with people in text will make pace of improvement much faster.  For the fun of it, here is Dropbox’s application

The latter is a compendium of winning short stories. Mysterious narrative structure, winking literary voice, and teenage boy utopia. Like a comic book.

But this frontispiece about his Adonis-like career and process is just so infuriating. I hate him.

Clues here https://discussions.apple.com/thread/8371206

and here https://wordpress.org/support/topic/only-10-episodes-are-showing/

Proof: https://podba.se/validate/?url=https://amol.sarva.co/feed/podcast/in-the-know

Here is a popular argument lately:
– there is one possible world where we are real humans with intelligence
– there are many many where we are simulated beings in some advanced civilization's AI.
– therefore it is probable that the "we are real!" feeling is just a simulation. Far more likely, counting the possibilities, that we are simulations not real
That's the Simulation Argument and it comes with a formula.

https://en.m.wikipedia.org/wiki/Nick_Bostrom <– will get you there.

I think it is wrong.

Mainly because the probability bit of it was already true. There are many many possible worlds where malevolent demon scenarios are why we feel we are real. Also dreams (all the dreams of all the people in all the nights of time, and dreams within dreams). And other forms of anti-reality.

But computer simulations add to this. So are there more?

That depends on whether many infinities are more than some infinities. Which is about Cantor Countability.

So simulations mean it is less probable that we are really in the real world.

HunterFarmer

B

Challenger/opinionated thought partner Relationshipper/friendProblem Solver/I will get the answersHard Worker/1,000 outbound per dayLone Wolf/don’t know how he does it

C

Money or Meaning

D

Run a machine or invent a machine

E

Top $ earner Or Top team player Or Top love earner 

F 

Jordan or Phil Jackson

There is an important theme people ignore, and I think it drives the long waves and short waves more than you guess.

Energy prices. The anchor LPs in the top large funds and in many of the mid-grade funds are sovereigns, along with pensions and endowments.

Endowments don’t really matter. The biggest is maybe 30bn.

Pensions do. Calpers and the other California pensions for workers are nearly a trillion together.

But the pensions have a high payout requirement to service their pensioners, around 3-5% per year. And a long-term insanely imbalanced liability for health care costs and imbalance from a younger workforce. Many folks think they don’t have enough money to meet their aspirations. They need impossible returns and impossible safety. And as a result they are also closely politcized and can rarely do cool stuff. Those are the US ones. There are some big ones in Japan (state managed, too) that are large. But rarely found in venture. Maybe in PE firms.

So the sovereigns. Five Middle Eastern groups have around 3T between them, Norway has 1T, Singapore has 1T, and there are a few more.

—more to say—-

But also address how interest rates modulate the search for return – lower rates means must go for some higher risk equity.

Before a great battle he slept in a chair with one foot on a table

He read Goethe, wrote love letters, short stories, and collected whole countries’ masterpieces into a museum he built to store them (now called the Louvre)

He loved classical architecture from the Greco-Roman era

He never said “an army marches on its stomach” but did obsess about shoes

He never won a battle at sea

He lost four(?) battles on land, one with ‘water’ in the name – and Acre, Russian campaign, and Lepzig.

He commissioned the grandest scientific expedition of his day, to learn about Egypt

He conquered Italy/Naples, Switzerland, Spain, Holland, “Germany”/“Poland”, Austria/Hungary (alliance), Denmark/Norway/Sweden (alliances), and Egypt. He also beat the Turks.

He never landed in England. Didn’t get Portugal. Didn’t win in Russia.

Nelson beat him at Trafalgar and died the same day

His masterpiece is considered Austerlitz, in Austria, where he beat the Russian and Austrian empires in a single battle

The strategy of the central position

L’attaque de la derriere

His hallmarks:

• be fast
• keep forces concentrated
• single language, single supply line, single communication corridor
• singular decision-making
• decentralized execution
• fearless of numerical odds
• divide the enemy
• fight where you have the advantage, even if the time is not of your choosing

Meet the troops themselves

Inspire the spirits of the troops

Keep a reserve, deploy it late

Engage early and wait for the decisive moment

Attack again after a victory

Page 1, Line 1 of the Rules of War: “Do not march on Moscow” (not in his book)

Wrong so far.

The Dow hit a high last September, but it looks like it will hit the next one very shortly.

When it does, well then I will be proven wrong.

Watch for 26,829. Maybe good times have more legs underneath them.

Here’s the book by Reckwitz, that I am very surprised seems not to have been written before.

A different book has been written before — the one by Marx, Horkheimer and Adorno, Walter Benjamin, and so forth. The one that says Ford and the Industrial Age changed culture into industrial product. Sameness everywhere. Grandmother’s recipe…at every corner restaurant. The signature fabric print of your ancestors’…for sale to every teenaged consumer.

Fast forward a few decades, and now we look back on these “product designers” and marvel at their creativity. They took control of Ford’s factories and churned out fresh aesthetic and functional variations on a stolid warhorse.

It went further. Capital, according to the Marxists, was stifling creativity. (Hopefully, that would undermine it, hence terms from the Baudrillard-types like “very, very late modern capitalism” to describe 1997.)

But capital struck back. Forget Henry Ford’s “any color you like as long as it’s black”. The world’s largest industrial colossus makes a pocket-sized gadget that every single one has a different baby picture on the home screen — Apple and the iPhone. Creativity at scale. A new one or two every year. Always different, never the same.

There’s more to it, and this book reaches back to the Romantics and Nietzsche to locate the origins of the creativity cult (übermensch has something to do with it). But I think the place to leave this topic is by noticing we are in the age of mass customization: Nike sneakers that say sweatshop, Starbucks orders just like all the others except with Amol written on the side, profile pages on social networks that let you plant a flag on the media empire, and advertising targeting that’s so personal you actually find it creepy sometimes.

So what’s next? Return to the impersonal, uncreative, cookie cutter? Or is the maker-made world the one that comes after the mass-customized one?

Nisbett is a legend of cognitive science and psychology. I think I cited him ten different ways in my dissertation back in the day. 

A popular recent book he wrote is “What is intelligence?” (the ability to figure things out), and there are his writings on education (a process that creates intelligence). If you have a quibble with either definition, I invite your emails. 

This book is about how intelligence might be different across cultures. People usually think “figuring things out” rests on an objective standard, so intelligence must too. What happens when you investigate?

You find some patterns that vary starkly between East Asian styles and Western styles. (I think he doesn’t really handle the Indian style.)

Before noting how they are different, doesn’t it immediately raise a few questions?

• what if intelligence tests cannot therefore be one-size fits all? and
• What if, inadvertently, the stuff that makes tests hard for Westerners is what makes them easy for Eastern tradition thinkers (children of Asian immigrants)?

On this latter point, Asians are 10% of NYC yet 70% of the top test-based high school. They are poor. They are in public school. Their parents do value study highly and do send their kids to cram schools — but what if it is more than that?

OK back to the book.

Key findings:

West vs East

Namely:

Conflict vs consensus. Westerners like debate. Not Asians.

Logic v context. X implies Y and so we must Z, say Westerners. In the East, the world is more accurately described when it is complicated and subtle. No one argument changes everything.

Clarity v nuance. Am I being clear? Vs: Am I being subtle?

Change v stability. West prefers action. East prefers continuity.

Words v silence. Smart people talk vs smart people listen (“one mouth, two ears”).

Consistency vs contradiction. A smart speaker is consistent internally, to the point of rejecting the received consensus. Newton or Galileo upend the old world. In the East however, the wise person knows “nothing ever changes, yet everything stays the same”. The contradictions in reality are reflected in the wise view of reality.

My mentor said: “How many of those are there? Just that one?”

I couldn’t think of any. I took out my cofounder shortly thereafter.

It is perhaps 10-20 years late but here we go:

Read my highlights from Powers of Two

Munger and Buffet

Magic and Bird

Pierre and Marie Curie

Lennon and McCartney

Abernathy and MLK

Susan B Anthony and Elizabeth Cady Stanton

De Beauvoire and Sartre

Marina Abramovic and Ulay

Crick and Watson

Jobs and Woz

Gates and Ballmer

Matisse and Picasso

Van Gogh and his brother Theo Van Gogh

Orville and Wilbur Wright

It continues. Of course this book is the first place I have ever seen an effort to collect all the great pairs, since most efforts are about the mythical Lone Genius.

Shenk wants to argue there arent any lone geniuses, but his strategy is an argument from induction (show that every single genius had a collaborator, one by one). He doesn’t get there in this book. (Tesla? Augustine? Philip Pettit (tightrope guy), Cezanne?) (Originally I had Nietszche on this list – but he got me, Salome, the muse.)

He does show that duos can he dynamic, and he expands the definition helpfully.

Your partner may not be forever, may be a rival or antagonist, a complement or a prompt. In all cases essential.

So here is the options framework for partnerships:

1. Star and Director. Both important, one famous.
2. Structureberg and Jokestein. Straight man/organized and color/punchline sharpshooter. Both important.
3. Inspiration and Perspiration. Dreams and Doings.
4. Turn-Taking. Trading the hot hand. Rewriting each other.
5. Yin and Yang. Opposites, complements.
6. The Other and Inner/Outer inspiration. Someone who helps you be you.

So what to do with this framework? I would suggest starting by labeling your partnerships. Which is which? Do they work as desired?

That cofounder I had to take out was selected by me to be the Star, so that I could help behind the scenes and push further. Turned out it had to be the other way around, with me as Star. I had never done it before. And I had to find a Director. Much later, I did.

Apparently founders resemble the hypomanic type.

This is the textbook type:

Perhaps advice books is not his forte.

I found one useful observation: the play is a quest for a solution.

Many artists, writers, comedians, and of course scientists share this aim as their orienting point: finding the truth.

But Mamet adds a helpful staging. The play starts when the problem has ripened, not too early. In this he adds a great insight. The 2 hours of dialogue won’t develop the content fast enough.

Painting is different perhaps? Maybe not.

Here is checkmate about to happen in a new painting by Tom Sanford:

BOOK 1  •  CHAPTER 6 

Information in War 

BY the word “Information,” we denote all the knowledge which we have of the enemy and his country; therefore, in fact, the foundation of all our ideas and actions. Let us just consider the nature of this foundation, its want of trustworthiness, its changefulness, and we shall soon feel what a dangerous edifice war is, how easily it may fall to pieces and bury us in its ruins. For although it is a maxim in all books that we should trust only certain information, that we must be always suspicious; that is only a miserable book-comfort, belonging to that description of knowledge in which writers of systems and compendiums take refuge for want of anything better. 

Great part of the information obtained in war is contradictory, a still greater part is false, and by far the greatest part is of a doubtful character. What is required of an officer is a certain power of discrimination, which only knowledge of men and things and good judgment can give. 

The law of probability must be his guide. This is not a trifling difficulty even in respect of the first plans, which can be formed in the chamber outside the real sphere of war; but it is enormously increased when in the thick of war itself one report follows hard upon the heels of another; it is then fortunate if these reports in contradicting each other, show a certain balance of probability, and thus themselves call forth a scrutiny. It is much worse for the inexperienced when accident does not render him this service, but one report supports another, confirms it, magnifies it, finishes off the picture with fresh touches of colour, until necessity in urgent haste forces from us a resolution which will soon be discovered to be folly, all those reports having been lies, exaggerations, errors, &c., &c. In a few words, most reports are false, and the timidity of men acts as a multiplier of lies and untruths. As a general rule every one is more inclined to lend credence to the bad than the good. 

Every one is inclined to magnify the bad in some measure, and although the alarms which are thus propagated, like the waves of the sea, subside into themselves, still, like them, without any apparent cause they rise again. Firm in reliance on his own better convictions, the chief must stand like a rock against which the sea breaks its fury in vain. 

The rôle is not easy; he who is not by nature of a buoyant disposition, or trained by experience in war, and matured in judgment, may let it be his rule to do violence to his own natural conviction by inclining from the side of fear to that of hope; only by that means will he be able to preserve his balance. 

This difficulty of seeing things correctly, which is one of the greatest frictions in war makes things appear quite different to what was expected. The impression of the senses is stronger than the force of the ideas resulting from methodical reflection, and this goes so far that no important undertaking was ever yet carried out without the Commander having to subdue new doubts in himself at the time of commencing the execution of his work. 

Ordinary men who follow the suggestions of others become, therefore, generally undecided on the spot; they think that they have found circumstances different to what they had expected, and this view gains strength by their again yielding to the suggestions of others. But even the man who has made his own plans when he comes to see things with his own eyes, will often think he has done wrong. 

Firm reliance on self must make him proof against the seeming pressure of the moment; his first conviction will in the end prove true, when the foreground scenery which fate has pushed on to the stage of war, with its accompaniments of terrific objects is drawn aside, and the horizon extended. This is one of the great chasms which separate conception from execution 

• Epicurus, Lucretius, Machiavelli, Mill means-end and utilitarianism.
• Aristotle, Kant, virtue and deontology as distinguished from the Stoics
• and do enjoy the stoics (basically Romans influenced by the Greek rule-follower types) like Seneca and Marcus Aurelius.

You could also just read this article about happiness.

It is neat to see tech elites getting teased for talking about philosophy. The one midwestern classicist guessing why such people are liking these ideas is a bit of a stretch. Basically: “These people are embracing Stoicism like the last time rich people talked philosophy, the romans”. (Did Rockefeller follow Marcus Aurelius? Did Kravis? Neither did the Romans: they also had their Neros and Caligulas.) Stoicisms ideas start long before the Romans, and were embodied by the penniless, street-living, hemlock-drinking gadfly types. Stoicism is certainly not a philosphy for the rich and powerful, primarily, and when such people talk virtue I would expect a more favorable reaction than of they talked hedonism. Oh well.

I think there is a simple reason tech people like stoicism: it is super stressful to start and build companies. Other things are stressful too out there. But tech is a bit more of a rollercoaster than the other rich-people professions like consulting or banking perhaps.

But the tech elites should also read a bit more widely. There is more great stuff to consider, like the sublime and the euphoric, or madness and eukrasia. The stoics don’t quite handle it all.

Anyway, here is some stoic news you can use (or epicurean, depending on what you want to pay this person):

It is about suicide. So yeah sleep is death.

Nas by contrast offers a steelier view of being awake:

Hamlet would have difficulty convincing me his situation is more grim than Nas above. But Nas rocks his Fortinbras stilo. “I think of crime.” So does Fortinbras near the end of Hamlet – he thinks of assasination.

Returning to it now as the new litigators of this topic like Coates and Cornell West have been tussling, and as new movements related to racial justice around incarceration, police violence, and reparations (did you see David Brooks get on board?) have been lively this past decade.

Dubois is one of the earlier American literary saints and thinkers on racial justice, and I finally had a look.

What an amazing and lyrical writer and historian. You could read this book just for its poetry of life in the South, or the saga of suffering under failed promises after slavery.

But I took a few particular learnings.

1. All the problems of today existed in 1900 — racial poverty, enfranchisement and voting, a shadow system of slavery enforced by unjust courts and police and local laws, a weak central government, a failure to amend the substantive and economic disadvantages left by a huge moral crime, a satisfaction with variations on formal equality. And Dubois’ point is actually that all those problems of 1900 existed in 1850 and 1800 also, in moderately different forms, under the full weight of American slavery. Yikes. Disappointment in the dream of freedom basically.
2. Black intellectuals didn’t agree with each other then either. Dubois’ goes after the “hey, just start a company and get rich like me!” philosophy of Booker T. Washington and all his peanut inventions. Painful to watch. Martin vs. Malcolm. (Or in post-Raj India, the conflict among early leaders like Jinnah and Nehru. It’s normal, I suppose. Solidarity is rarely durable. Hamilton and Burr…?)
3. The specific history of the whole thing — the pullbacks and manipulations of the Amendments, Reconstruction, the short-lived Freedmen’s Bureau, and all that. Way more than I recall from my 1 day of Reconstruction in 11th grade history class, pretty much all I learned about it. 40 Acres, Jim Crowe, the formation of the white supremacist groups, the reign of terror on black people before and long after Emancipation, etc etc etc.

That’s it for my tl;dr. Have a look at the link for some passages that got me.

Who highlights

Reading Who by Geoff Smart

17 new photos added to shared album

Old me would have luxuriated in the pleasure of crushing the airhead authors and their pamphlet-of-tips-dressed-into-Hire-Us-the-Book form.

I will instead summarize the argument of the book:

1. Know exactly what you are looking for in extreme detail when hiring
2. Pursue the absolute best people and don’t settle for anything lower than the best best best
3. One and two may take a little more time than all possible alternatives but don’t worry, and no we didn’t attempt to measure or quantify that
4. Because…the best do better than the rest!
5. QED

I will not bother disagreeing with vapidity of this chain of reasoning. Any big shot CEO who thinks there is nothing more to it than this…well they have more time and money than me.

It also seems to be a book for establishment companies that are not trying to change very much.

Now, there is something good from getting a couple of HR consultants to sit in front of a word processor and type for a couple weeks.

They have their list of questions.

Good questions are super super handy. I have this theory of Golden Questions. You can ask 200 questions in a satisfaction survey but only one truly matters (net promoter…).

In many areas this is true. In a roomful of high scoring grads of elite schools, why ask a puzzle question? Ask maybe “What’s the single most difficult thing you have ever accomplished?” Or propose a walk and ask about their favorite building in New York.

You can read their questions at that link of highlights.

Au revoir, Who!

The Yahoo Peanut Butter Memo

The Nokia Burning Platform

Can’t think of a famous defensive pivot.

Apple came back from the dead.

Microsoft retreated on consumer.

Facebook made the pivot to mobile.

But when Facebook today is pivoting to privacy they aren’t surfing a innovation wave. They are on a burning platform, with a peanut butter strategy.

Lifetime highlight. A note from my idol on a topic I care a lot about with the gesture that maybe he cares what I think!

More of my relevant writings here

The word “diversity” appears zero times.

Here is Google’s Diversity Report from 2014, when they *started* reporting it as a company (or perhaps started measuring it).

And here it is in 2018, four years later. In percentage terms some groups expanded quite a bit — lots more Asians, African Americans, Latinos.

No more women.

And by “lots”, well take a look.

30% Asian going to 36% is both a large proportionate increase and a large number of people –> 6% of Google’s staff is net-new Asian share.