Coté

Thu, 28 May 2026 10:36:38 +0200

Garbage Chairs of Amsterdam

Good O-Rings and Bad O-Rings, the AI Efficiency Plateau, and Anti-Labor by Design - Related to your interests, Thursday

Thu, 28 May 2026 10:11:26 +0200

Also: Goldman’s 24x token forecast, Indian IT’s process-debt pitch, NatWest’s 35% AI-generated code, Gartner’s 84% productivity theater, vibesec.

From: "AI Agents Forecast to Boost Tech Cash Flow as Usage Soars," Goldman, May, 2026

‘I’m delighted to ⁠be wrong’: Sam Altman says AI won’t lead to a ‘jobs apocalypse’ - but admits he was ‘pretty wrong’ on the social and economic implications it is having - “It really, in both positive and negative ways, updated me to thinking that the jobs picture is likely to be very different than we thought,” he said. CEO Said a Thing: “I don’t think we’re going to have the kind of jobs apocalypse that some of the companies in our space advocate or talk about.”
The AI efficiency plateau - Yes: “Among developers who reached the highest time-savings band, roughly 7 in 10 (69.7%) got there in less than two quarters.” But: “Of those developers who reached peak time savings, two-thirds (66.1%) reported lower time savings in the quarters that followed.”
AI Agents Forecast to Boost Tech Cash Flow as Usage Soars - Yes: “The important point is that the adoption rates are still relatively low today, especially in small to medium-sized businesses. In 2030, we forecast that 12% of knowledge workers will be using agentic AI yet by 2040 that figure will be 37%. You have this very long tail adoption.” But: “Agentic AI is expected to drive a 24-fold increase in token consumption by 2030 as consumers and enterprises adopt the technology, according to Goldman Sachs Research” // The vibe I get is that a lot of that token increase is consumer and end-user oriented: replacing search, organizing inboxes.
U.S. companies have an AI problem. Indian IT wants to be the solution - ‘“The real question in enterprise AI is not who builds the most capable model. It is, ‘Who can make AI work inside messy, complex enterprise environments that have accumulated decades of process debt, data debt, technology debt, and cultural debt?'” he said. “That is precisely the terrain Indian IT firms know best."’
AI at work is anti-labor by design - Their theory: enterprise AI ROI is laying off people.
Gartner Says CFOs Risk Falling Behind Without a Scalable AI Strategy - Excerpts here.
NatWest inks AI deal for trade finance - “Headline figures for 2025 saw the bank’s software engineers generate 35% of its code through AI software development tools, all 60,000 staff given access to AI productivity software, and thousands of human hours saved. Last year, the bank also embarked on a major collaboration with AI supplier OpenAI.”
Trisha Gee: AI Amplifies What’s Already Broken

Found by Alan Jacobs.

AI Summaries

I wanted to read these, but I didn’t make the time, so I asked the robot to summarize them.

Found by Alan Jacobs.

Wastebook

vibesec.

ICYMI

A short video from me: When it comes to enterprise AI, here’s three things I’ve been hearing: (1) cost - we finally have to pay for this stuff - holy cow! (2) Cases - programming is great, but what else can we use this for aside from customer service and gussied up search? (3) Control - we can’t even track costs, how are we going to manage everything else?
Elusive Enterprise AI ROI: No scaling, it’s not legible, lack of skills/need for training - there were several things saying AI for ROI is not doing as great as planned. Here, I round them up.
Enterprise AI Slop
The O-Ring and the Keystone: Two Readings of Where Humans Sit in an Automated Economy - In the AI replacing humans talk, you hear about O-Ring theory sometimes. It goes something like this: in a ten step process, if you mess up one of the steps, even if the other 9 are perfect, the whole process is tanked. The positive reading is something more like: you don’t automate aggressively unless you trust the O-ring. // Anyhow, here is AI writing it up for reference and connecting to AI and other jobs.

Want to subscribe to this newsletter and get it in your email? Do that here. You’ll just get this type of link and post round-up, not everything posted on the weblog.

Elusive Enterprise AI ROI: No scaling, it's not legible, lack of skills/need for training

Thu, 28 May 2026 07:03:40 +0200

Despite everything, reports are still that enterprise AI ROI is elusive. At the same time, for enterprise buyers, the bill is finally coming due for the past year of AI amazement. It’s not cheap.

What’s up with this elusive enterprise AI? Gartner has some survey-driven theories for finance departments.

One theory is that there actually are productivity gains but they accrue to individuals, not the organization:

Gartner data presented in the keynote indicated that 84% of finance AI spend relates to individual productivity and process improvement use cases, while only 16% goes toward use cases that can materially change business outcomes.

Enterprise ROI does not scale, at least you can’t measure it. The leftist way of looking at it is that labor captures the value, not management/capital. For a human, this is great! Or, if you ponder at layoffs attributed to AI, it’s the opposite!

This, if you can’t measure it, you can’t ROI it:

According to the keynote, 71% of typical finance teams report low impact from their AI investments, and 62% of CFOs say fewer than a [25%] of their AI initiatives deliver measurable benefits.

I suspect this lack of seeing “impact” is also because we don’t know the uses of enterprise AI yet. Low in my list is that AI doesn’t actually work that well. It works well for me!

That said, like any other new tool, people need to learn how to use it. Typically, organizations are terrible at training people:

A major barrier is talent. Gartner research shows that only about 30% of finance talent currently qualifies as digital talent – employees who can build a technology solution when they encounter a problem – while breakaway firms are targeting 90% or more.

This has been the case for every type of new technology - remember all those Kubernetes surveys that showed that lack of skills was a top three problem? I’m sure relational databases and ERP systems had similar survey results.

Now, this all scoped to finance departments. However, I believe that it’s safe to generalize to other roles.

🔗 Gartner Says CFOs Risk Falling Behind Without a Scalable AI Strategy

CEO Said a Thing: jobs apocalypse

Thu, 28 May 2026 06:34:35 +0200

It really, in both positive and negative ways, updated me to thinking that the jobs picture is likely to be very different than we thought. I don’t think we’re going to have the kind of jobs apocalypse that some of the companies in our space advocate or talk about." Sam Altman

From: ‘I’m delighted to ⁠be wrong’: Sam Altman says AI won’t lead to a ‘jobs apocalypse’ - but admits he was ‘pretty wrong’ on the social and economic implications it is having

Related, the case for not believing anything the billionaires say, which is to say, the case for them not talking.

Enterprise AI Slop

Thu, 28 May 2026 06:27:08 +0200

People are using AI to generate too much work because they think they know what they’re doing:

A growing body of work calls this output-competence decoupling. In any previous era, the quality of a piece of work was a more or less reliable signal of the competence of the person who produced it. A novice essay read like a novice essay; novice code crashed in novice ways. AI has severed that relationship. A novice now produces work that does not betray the novice, because the competence the work reflects is not the novice’s competence at all. It is the system’s. The person, in the transaction, becomes a kind of conduit, capable of routing the output to a recipient and incapable of evaluating it on the way through."

There should be a best practice with AI, always ask it to make the output shorter:

Requirements documents that were once a page are now twelve. Status updates that were once three sentences are now bulleted summaries of bulleted summaries. Retrospective notes, post-incident reports, design memos, kickoff decks: every artifact that can be elongated is, by people who do not read what they produce, for readers who do not read what they receive. The cost of producing a document has fallen to nearly zero; the cost of reading one has not, and is in fact rising, because the reader must now sift the synthetic context for whatever the document was originally about.

From: Appearing Productive in The Workplace.

CEO AI Psychosis, Information Cleaners, and Mid-century Architecture Cartoons - Related to your interests, Wednesday

Wed, 27 May 2026 09:34:18 +0200

Also: Zoom system-of-action, the curl deluge, the unreasonable effectiveness of HTML, behavior-first mainframes, fewer books, the meeting Ask, and the multi-agent delegation problem.

Executives have a narrow view of AI gains - “CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI.” Aaron Levie Box CEO.
The users have plenty of feature ideas - in contrast, in favor of doing the whole “citizen developer” thing.
The Copy and the Guru - Use AI to improve how you think and work, not replace and multiply you.
Using Claude Code: The unreasonable effectiveness of HTML - Why and when you should use use HTML over markdown as the way of working with AI.
Google I/O: Did Google Just Ship the Full AI Stack? - “The governance layer is the part of the stack that the keynote narrative left unclear…. It is split across the Antigravity IDE and Agent Platform, and the announcement narrative did not unify the picture.”
Who Authorized That? The Delegation Problem in Multi-Agent AI - People are obsessed with responsibility in agentic AI. Part of this is driven by regulations, where we want to know who to blame when things go wrong. Also, attribution of good results would be cool too.
The pressure - “The rate of incoming security reports [for curl] is 4-5 times higher than it was in 2024 and double the speed of 2025 - meaning that on average we now get more than one report per day.”
The Behavior-First Paradigm: Moving Mainframe Modernization Past LLM Wishful Thinking - No one remembers how these things work, or what they were supposed to do. Maybe the AIs can figure it out by observing what they do.
The Ask - Figuring out what the point of a meeting is when the Americans won’t just tell you directly.
Zoom’s most recent quarter highlights its transition to a system-of-action company - “Zoom outlined its vision to become a ‘system of action.'” // They’re great at video conferencing, so I hope they keep doing that well.
Information Cleaner - Removing ads and paywalls as a social good: “An Information Cleaner is a person who takes in all the material being published in our information atmosphere and cleanses it: they make it transformable, searchable, modifiable, accessible, free of ads and trackers, auditable, connected to other information where relevant, and so on.”
The Cartoonist Who Mocked the Madness of Modernism - Who knew you could have architecture cartoons?
Usually life hacks and productivity tips are the same old things, these are fresher
Why are you reading fewer books? - “New books are appearing all of the time. Older books are still around. But if you are spending less time with books and more time with other media, that does not mean that you are becoming illiterate. Chances are, you are becoming more discerning about how you use your time.”

ICYMI

How do AI Layoffs Work? Some Speculation. - when you hear that thousands have been fired and replaced with AI, what exactly is the AI doing?
The AI Security freak-out: now is the time for platform engineering to shine - if you have a good platform in place, you’re in luck!
trying your best properly rated

Conferences

Conferences I’ll be at and some that I’m interested in.

J-Spring, June 4th, 2026 in Utrecht - Coté speaking.
WeAreDevelopers Europe, July 8-10, 2026 in Berlin, Coté speaking.
WeAreDevelopers NA, Sept 23-25, 2026, Discount Code: DEVPOD26.
DevOpsDays Istanbul, October 24th, 2026 - Coté keynoting.
VMware User Groups (VMUGs) - Dallas (June 9-11, 2026), Orlando (October 20-22, 2026).

Logoff

Record temperatures in Europe - at least it feels like that.

Want to subscribe to this newsletter and get it in your email? Do that here. You’ll just get this type of link and post round-up, not everything posted on the weblog.

Usually life hacks and productivity tips are the same old things, these are fresher

Wed, 27 May 2026 06:53:50 +0200

Stop brainstorming. There’s a mountain of research showing that team brainstorming doesn’t work. There are several reasons why, some having to do with unconscious conformity, others with fear of being judged, and still others with unclear norms. (For example, people are often told to say whatever comes to mind, but also not to criticize.)

Some six page memo vibes there. Come to the meeting with a proposal. And, for as much as I loath the pre-wire and socializing the deck, those does give people a chance to consider an idea. The alternative is that they learn about it for the first time in a meeting and then have just a few minutes to evaluate and react. The problem is more the medium of the meeting, not collaboration among workers.

Here’s a summary of the other techniques and practices mentioned:

Monotask. Pick one important task, set a one-hour timer, and put your phone in another room. Within days this retrains attention that has been conditioned to switch constantly, and two or three blocks tend to beat a full normal workday.
Satisfice instead of maximize. Set “good enough” criteria in advance and stop deciding once they’re met. Maximizers end up less happy and more prone to regret, and most daily decisions don’t repay the cognitive cost of optimizing.
Brainwrite, don’t brainstorm. Have everyone write ideas independently before the group discussion. It defuses conformity and fear of judgment, and produces the equal turn-taking that Carnegie Mellon, MIT, and Google research links to the smartest teams.
Build shared obligations. Tie family and community life to real recurring duties, including chores for kids. The Harvard Study of Adult Development’s 86-year data shows strong real-world ties are the single best predictor of health and longevity.
Use deadlines and commitment devices. Put a date on it, pledge money to charity if you fail, or otherwise design the environment so the desired behavior is the default. This replaces willpower with structure and makes bailing psychologically or financially expensive.

See more details in this longer summary of the summary.

And, check out the original: The Counterintuitive Way To Get Better At Anything by Eric Barker.

The coming AI backlash: capital is not here to make friends

Wed, 27 May 2026 06:45:37 +0200

[O]ur fears of what AI will do to us are really just our fears of what capitalism is already doing.

And:

the threat isn’t so much that AI is inevitable as that the ongoing—and likely expanding—immiseration of workers is unstoppable. This is the subtext of the strange and conflicted messaging that we get from the hype men: when they say that you better learn AI or be left behind, they are admitting that a great many people will be left behind. And if you—smart and clever and hardworking person that you are—are somehow able to make it to the other side of the line, you’re supposed to find relief or pride at having done so, and not horror at all the people suffering in your wake. You’re supposed to be as uncaring as the capital that uses you.

From “Why We Fear AI."

How do AI Layoffs Work? Some Speculation.

Tue, 26 May 2026 12:54:57 +0200

Via geralt on Pixabay

When an executives says layoffs were driven by AI, what exactly is the AI doing that removes the need for those humans?

Here’s some dog-walk speculation.

Decks, Meetings, etc.

All the prep work around The Meeting. Things like: the agenda, slides, the pre-read, notes during the meeting, and followup tracking. There’s the careful synthesis of who said what so it can be presented in a different room to a different set of people for the next round of synthesis. In any sufficiently large organization this is - what? - 10% to 20% of knowledge-worker time? Maybe more.

Maybe when companies lay off people, they are relying on AI to do a lot of this work. Not perfectly, but maybe good enough versus paying teams of people to do it. Or maybe AI does it even better than humans!

You can imagine an SVP with no VPs and just a handful of individual contributors. All of the corporate data is connected to a chat application and the SVP just talks with the AI. They might then have a brief meeting with the handful of people there to discuss what they think and, of course, what their AI chat sessions said. There’s probably some MCP Servers. You could imagine some markdown files (skills and plugins in the Claude Cowork sense) forming the foundation of it.

Contra to this is the notion that the deck is going to be skimmed for 90 seconds anyway. The “real” work is done by people informally talking with each other to “pre-wire” and “socialize” the deck. But, maybe you throw in some Agent-2-Agent to have your AI talk to their AI(s).

The measuring middle

“The vast majority of those we laid off last week were measurers,” he wrote. He defined “measurers” as those in middle management, finance, legal, internal auditing, and revenue recognition. Matthew Prince, Cloudflare CEO

It turns out, executives who lay off staff tell us, there’s a lot of people who just measure and track things. Part of this is getting the deck together and doing The Meeting, above.

What does the measuring middle do? Figuring out what to measure, then measuring it, then talking to the people being measured about those measurements. Daily status. Quarterly briefs. Career conversations. Maybe even the board deck! The annual strategy and budgeting planning. Analysis about new businesses to enter, etc. Each year, of course, you need to revisit how you measure. And when a new executive comes in, you need to introduce a new measuring system and then take the time how to apply the new measuring system to how you’re working.

Why pull six VPs into a strategy offsite when each board member can just talk to Claude about the state of the business and get a tailored brief? And, of course, each of those six VPs has teams of people who get their deck and talking points ready, and then respond to the other VPs.

AI can do most of that, the Cloudflare CEO says:

Tireless, independent, efficient and available, AI systems can now measure an organization with a level of objective detail and precision that was previously impossible even for the best employees.

This feels like where most of the AI-attributed tech layoffs actually land. Maybe? Project managers, program managers, chiefs-of-staff, the layer of “I run the cadence.”

Reporting

AIs can probably do all those dashboards and detailed reports that need to be executed into short emails. This adjacent to the meeting stuff but distinct. This is the long-promised thing of actually knowing what is happening in your business day to day, not just project-status.

Real “what changed this week and what does it mean” reporting that the BI team has been promising since 2008. AI is finally making that cheap enough to do once you hook up chat apps to all those data lakes.

Operations and paperwork

Anyone who has filed a tax return knows how much of organizational life is filling out a PDF and routing it somewhere. Corporate life is full of filling out PDFs (and, even more horrifically for some, ERP web forms): POs, invoices, the paperwork to open or close a bank branch, the corporate-law boilerplate for every non-trivial action a company takes. One could imaging SEC files. Have the AI do it. Have a person review it. Adjacent: the lower tiers (or upper!) of corporate law, which is a lot of “here’s the standard M&A doc, change these seven fields.” There must be a lot of “prepare the paper work to send to the person/organization” that goes on.

Corporate communications

When you need to tell the people in the organization something, instead of spending time with humans to come up with what to say and how, and then to actually say it in various mediums…just have an AI do it all.

Again, an executive sits at an AI chat app and talks with it, finally saying, “that looks good, send an email to all employees with this exciting update.”

You could even train on the executive’s voice and visual and send out an AI generated video.

This could likely be extended to all communications (see marketing and design below).

Programming

Programmers are well covered. But it also includes architects, QA, project managers and measurers, product managers who decide what features and fixes get in, security people, even ops people who are monitoring and fixing problems in production.

The executive just logs into the chat app and says “hey, how are the servers?” Or, “hey, can you add the ability to sell our tires in Cyprus?”

New AI features

The business adds new features to its existing software and, thus, business. We saw a lot of this driven by humans during COVID where banks had to do new types of loan applications, retailers had to support curb-side pickup and return, restaurants needed scheduling and COVID paperwork checking, etc., etc. Humans did this insanely fast - with the help of lot of “digital transformation” of how they did software and platforms they used to run apps.

Now, imagine if new features were instead driven by an executive’s imagination and they could just sit in front of their AI chat app and say “wouldn’t it be cool if we had a new type of corn dog?”

Via geralt on Pixabay

Customer service

Mostly automatable, with an asterisk. In my experience, the reason humans were ever required for customer service is that eventually a decision has to be made or an action has to be taken. The old “automated customer service” felt like just keyword search against a knowledge base, which is why everyone hated it and immediately typed “agent” eighteen times into the chat. When you call customer service, it’s usually because you want something to happen, not just learn about how the business functions and its thoughts on you. For example, you want to cancel a credit card, change a flight, etc.

The new version can actually do the thing - issue the refund, reschedule the order, close the account - and that changes the math. I mean, assuming you trust its judgement - but the cost of the errors might be much less than the cost of humans, and speeding up the process might include customer service that, you know, a few first class ticket given out here and there is no big deal.

Diagnostics

From the mechanic figuring out why a jet engine is acting weird to the platform engineer figuring out which of the seventeen services in the request path is causing the latency spike in equity trading. Maybe AI is good at “given all this state, where is the problem.” Hopefully.

Planning

An executive sits down in front of an AI chat app and says: how much ground beef should we order for next quarter? How much steel? How many people should staff the night shift? Should we open that new branch in Cleveland?

This used to be spreadsheets plus a twenty-year veteran’s pattern-matching. You want ground beef to Rotterdam next quarter, not a kilo more or less. But, it’s August, so you know even though they say they can handle a tripling in ground beef, you know the dock workers might mostly be on vacation.

Just unzip a few markdown files into .claude, OAuth to that ERP system, and ship that meat. No need for so many humans.

Marketing

There’s also marketing, design, graphic stuff - why go through the trouble of generating copy, infographics, brochures, and even signs? It’s not great at writing, but if you just want to get a point across quickly, or explain something, it’s definitely cheaper and faster than a human.

Instead of hiring professionals, I use AI a lot for video and other content production. Descript is amazing for this kind of thing and it’s loaded with AI.

I was in León recently and noticed that even the churches there were using art from ChatGPT (the old anime looking stuff), and a lot of signage at restaurants too.

In general: one person doing the work of three

One person can now do the work of 3+ people. Each employee can now get things done faster (doing one or more of the above), so they can use the 40 hours a week to do more work. This is “productivity”: you are paying less for the same output. You can also pay less and/or the same amount for even more output. That’s kind of like “productivity plus”!

If one person can do the work of 3 people, you either get rid of two people if you want to take on the risk of (a) there is no new work, (b) that person stays and has good enough availability between sickness, vacation, and leaving the job…or you get rid of one person, retaining a “backup human.”

Maybe the AI is the “backup human” too!

What this list has in common

The categories above are almost all white-collar coordination work. Meetings, measuring, reporting, paperwork, communicating, deciding what to decide. The layer of work that grew enormously over the last thirty years as companies got bigger, more matrixed, and more globally distributed. Work whose product is a document another human reads so they can produce a document another human reads.

The classic comparison is one of those black and white pictures of a bunch of clerks in an office that are replaced by a spreadsheet.

Some people like to imagine that AI cannot reduce “manual” work, from plumbing to nursing, radiology, and doctoring. But, you’ve got robots (one day, we’re told - I saw last week that someone has made advances in folding clothes), and if you can take a picture of a leaky pipe and have AI tell you how to replace it, maybe less plumbers too. Of course, getting a plumber to come instantly is often difficult, so maybe we don’t have enough plumber supply. Until robots advance, I think you still need humans to build houses, cook food, cut your hair (definitely straight razor your neck!), etc.

Via geralt on Pixabay

This meeting could have been a chat session

To day-dream more, think about any case where you’d say “this meeting could have been an email,” and that’s likely a good candidate. That’s a start, but there’s clearly there’s lot more.

What have you come across?

The AI Security freak-out: now is the time for platform engineering to shine

Tue, 26 May 2026 09:14:07 +0200

The AI-driven security freak-out is a time to see what if your platform engineering strategy is working. A good platform makes it possible - if not easy - to find and patch all these new CVEs. And, of course keep patching them.

A good platform will keep track of all these apps and dependencies deployed on the platform; be able to rebuild apps and services with minimal to no developer work; and be able to roll-out upgrades, rotate keys, and otherwise “seamlessly” deploy the patches. Plus all the reporting and audit stuff.

Check out Darin’s overview for more and how Tanzu Platform does it. An excerpt:

The security landscape just shifted under our feet – again. Over the last 18 months, AI-assisted vulnerability discovery has compressed the timeline from novel CVE published to weaponized exploit in the wild from weeks down to hours. Researchers (and bad actors) are now using LLMs to chain together previously unrelated weaknesses into novel zero-day attack paths. The volume of disclosed vulnerabilities keeps climbing, and the half-life of “unpatched, but probably fine” is collapsing.

For platform engineers, this means the old rhythm of quarterly patch windows, hand-rolled CVE spreadsheets and “we’ll get to it after the next release,” is no longer a defensible posture. The only durable answer is the boring one: Rapidly apply first-party, vendor-supplied, vendor-supported security fixes across the entire estate before the chained exploit lands in your environment.

The problem isn’t whether to patch. It’s knowing what to patch, where it lives, and how to quickly roll it out without disrupting the business. With VMware Tanzu Platform 10.4, this can all be done across your entire Tanzu Platform foundation fleet, in a single workflow.

We also have some new advice for management types who are looking to develop and evolve their security strategy.

If you find out that you don’t have this all in hand, better TryTanzu.ai.

AI Pees in the Pool, Permanent Indenture, and the Metric Fuck-Ton - Related to your interests, Monday

Mon, 25 May 2026 09:21:29 +0200

Also: does anyone actually use microwave food-buttons?

Castanet handler.

Platform Engineering in the Age of AI: Why Operational Complexity Is the New Bottleneck - “while coding is becoming cheaper and faster, operational complexity hasn’t disappeared” // Day 2 operations is always the bottleneck. // Also included is the current Syntasso/Kratix pitch.
🤖 Five takeaways from Michael Dell’s keynote at Dell Technologies World 2026 - “Dell’s most direct strategic move was to counter the idea that AI belongs primarily in the public cloud. Citing Dell research, Michael Dell said that ‘67% of AI workloads already run outside the cloud’ and that ‘88% of respondents are running at least one AI workload on prem,’ arguing that ‘CIOs are aggressively pivoting to hybrid AI.'” // Summary
Gartner Forecasts Worldwide AI Spending to Grow 47% in 2026 - “Worldwide spending on AI is forecast to total $2.59 trillion in 2026, a 47% increase year-over-year, according to Gartner” // At that level, it’s kind of incomprehensible, so just file it away as “a metric fuck-ton.”
Gartner Survey Finds Sales Organizations That Provide AI-Enabled Next Best Actions Are 2.6x More Likely to Achieve Commercial Growth - For better AI ROI: (1) train staff on how it works, (2) have the AI suggest what to do next, i.e., drive a decision, not just inform, search, and context.
You Need AI That Reduces Maintenance Costs - “I’ll get straight to the point: your AI coding agent, the one you use to write code, needs to reduce your maintenance costs. Not by a little bit, either. You write code twice as quick now? Better hope you’ve halved your maintenance costs. Three times as productive? One third the maintenance costs. Otherwise, you’re screwed. You’re trading a temporary speed boost for permanent indenture.”
Just aim the cannon correctly - Use LLMs to automate toil and low hanging fruit in the dev cycle. Adding features is fine, but it brings all the extra work and long term maintenance with it.
Prompts are technical debt too
1Password extends OpenAI collaboration with Codex MCP server for just-in-time credential access - I’m thinking this is a general MCP server. Well, hoping.
Laying it on thick - “Besides being plainly rude and dishonest, these messages ‘pee in the pool’ of internet communication, making it more difficult for sincere creators to send authentic emails about their projects, simply by raising the ‘noise floor’ of simulation and bullshit.”
LLMs and Buttondown - AI driving a SaaS funnel, here the email/newsletter service buttondown. Also, an API is an important part. Unlike substack, buttondown has an extensive API. Because of that, I keep eyeing it for my own newsletter list.
The Identity Crisis AI Didn’t Warn You About - “For most of modern history, we’ve invested perhaps 10% of our energy in ‘be’ and 90% in ‘do’. The AI era invites (demands?) something closer to the reverse. That’s not a comfortable shift for an industry that prides itself on shipping, but might be the most important thing we build next.”
AI Rage Is Inextricably Fused With Justified Loathing Of The Extraction Class. ‘Deal With It’ - From the read the room/Picard face-palm files.
Liberty Mutual, Travelers discuss when not to modernize - “We’ve got 70% of our compute in the cloud.” // “‘You want to modernize what’s in the critical path,’ said Mojgan Lefebvre, EVP and chief technology and operations officer at Travelers, speaking on the same panel. ‘We’ve chosen to not modernize some things and actually isolated subsystems behind APIs and then invested around it where the customer value or productivity was.'”
Nobody is destined for greatness.
Tyler Cowen: Wokeness Has Peaked. What Followed Is Worse. - This reads like the ultimate Cowen-Straussian piece, i.e., something like: look at us now, with the world on fire…oh to be back in the days when all we debated were pronouns.
What matters - “widely accepted reading of the-west-wing, as such, is that of a liberal Fantasia, not because it depicts the Bartlet administration as being liberal, but because it depicts a world in which the best possible thing you can do is to try very hard to do the right thing - and the liberals, or really the centrist-leaning Democrats, are the ones who do that.”

Wastebook

“synthetic text extruders” world war web.
“Then someone discovers firmware in a soldering iron and the whole IoT farce comes back into focus.” Here.
“My most successful writing comes from when I have a breadcrumb into a maze that I can follow to its logical conclusion” maelstrom.
“Humans are supposed to just walk past an aisle of candy bars?” Wegovy
“My family. My friends. Fun.”
Does anyone use those food-based buttons on microwaves and air fryers? I never do. A knob of power/degrees and then a knob of how long is always faster and reduces cognitive load.

AI Summaries

I wanted to get the ideas here, but did not want to read them. So the robot did.

ICYMI

How many quadrillions in a Googol? - Software Defined Talk #573 - “This week, we discuss Google I/O, the OpenAI soap opera, and ChatGPT going full financial advisor. Plus, thoughts on improving the conference hallway track.”
trying your best properly rated
AI backlash, from the read the room/Picard face-palm files.

Logoff

That’s a big build-up above.

Want to subscribe to this newsletter and get it in your email? Do that here. You’ll just get this type of link and post round-up, not everything posted on the weblog.

trying your best properly rated

Mon, 25 May 2026 08:10:18 +0200

widely accepted reading of the-west-wing, as such, is that of a liberal Fantasia, not because it depicts the Bartlet administration as being liberal, but because it depicts a world in which the best possible thing you can do is to try very hard to do the right thing — and the liberals, or really the centrist-leaning Democrats, are the ones who do that.

It’s a cliché, but that is a notion I haven’t encountered in a long time. In software, there was a time when design was in the ascendancy and we were advised that paying attention to what the user needed and wanted was key. (That is poorly worded.)

Of late, the focus has been on scale, productivity, whatever Kubernetes is, and now AI. It is about the potential to do things, pure motion and ability. But is it trying our hardest to do the best thing?

No. It never starts with “the best thing.” It starts with either finding a new way to make money off people’s attention, occasionally giving them a new “toy,” and of later productivity for productivities sake.

We don’t really need AI and code generation. We were doing just fine before all of that.

And, of course, the unspoken best thing is for the employee, founder, investor, etc. to make money.

The best thing for the user is whatever makes me the most money is the unspoken ethos of tech.

🔗 What matters

The users have plenty of feature ideas

Mon, 25 May 2026 08:05:27 +0200

AI allows people who aren’t software engineers to build meaningful software. Those of us who are software engineers at companies should stop building features and focus instead on building systems that allow people on the sales team, the factory shop floor, etc. etc. etc. to ship safely.

🔗 notes from o11ycon 2026

AI backlash, from the read the room/Picard face-palm files.

Thu, 21 May 2026 06:28:54 +0200

The normies don’t like AI, esp. The Kids:

They’re being told, usually by people who already have theirs, that they should be more excited about the latest evolutions in software automation. “Why aren’t you more interested in nuanced conversations about the latest evolutions in software automation,” fans of the latest evolutions in software automation will ask.

And:

I’m sure we’re due for a few Atlantic and New York Times brunchlord opinion columns condemning young graduates for not being appreciative enough of consumer-culture innovation.

Not only does the tech take away their cash before they already have, but the people doing are casually throwing off weird-to-reprehensible vibes:

They’ve watched tech oligarchs spend the last decade mired in scandal after scandal, hype cycle after hype cycle, steadily enshittifying everything they touch along the way. They just got done watching our top tech titans _eagerly_cozy up to the most virulently racist, corrupt bastards U.S. government has ever seen.

🔗 AI Rage Is Inextricably Fused With Justified Loathing Of The Extraction Class. ‘Deal With It’

Do less

Thu, 21 May 2026 05:49:14 +0200

For most of modern history, we’ve invested perhaps 10% of our energy in ‘be’ and 90% in ‘do’. The AI era invites (demands?) something closer to the reverse. That’s not a comfortable shift for an industry that prides itself on shipping, but might be the most important thing we build next.

🔗 The Identity Crisis AI Didn’t Warn You About

Now you can react faster than ever to security problems

Wed, 20 May 2026 15:21:40 +0200

This is an excerpt from our Tanzu Catsup last week. In that episode we talked all about how this AI stuff is changing - for the better - how you can handle security problems at the app layer.

It’s Monday morning. Your boss walks up, says “scrap the backlog, we’ve got a list of CVEs longer than that curved screen we bought you last year, the CISO is coming, fix them,” and goes to brunch. A year ago you had 30 days to triage anything under a CVSS seven. Now two lows can chain into a nine, the exploit is already in the wild before policy catches up, and both sides of the fence have AI for finding more.

This is the 3-minute version of ‪@dashaun‬ joining ‪@thecote‬ and David Zendzian on Tanzu Catsup to walk through what that Monday actually looks like at the application developer level. DaShaun on the new patching math and OpenRewrite. David on what AI plus an MCP server does to a pen-test timeline. The metric question at the end.

Check out the full episode, it was fun and packed with good discussion.

Shadow AI Surge, The Coming AI Backlash, and Apollonian Tyranny - Related to your interests, Wednesday

Wed, 20 May 2026 10:05:32 +0200

Also: Claude Managed Agents, Antigravity, Project Glasswing, repugnant economics, and writing observations.

From: Wes Anderson's Impossible Dreams

Innovations from Google I/O 26 on Google Cloud - Google seems the cleanest when it comes to describing their AI stuff…it’s something like a mix of clear naming (aside from “Antigravity”) and functionality, and not carpet combing you with tools like AWS.
Announcing Claude Managed Agents on Cloudflare - “Integrating with Claude Managed Agents is another step in this direction. You can run your agent loop on the Claude Platform, while using Cloudflare to execute code, secure connections, and run custom tool calls.” // Multi-cloud, hybrid-cloud, etc.
🤖 Agent Skills Deliver Gains Only When Curated, Structured, and Secured - Structure workflow work best with skill, and AI is not great at writing skills. // Summary
Project Glasswing: what Mythos showed us - Excellent overview.
Shadow AI surges in the workplace - “Of the 45 percent of all professionals using AI in the workplace regularly, 67 percent of those were accessing the platforms using personal accounts that were not authorized by their IT teams, data from Verizon’s annual data breach investigations report showed. Verizon said that the proportion of users accessing AI through personal accounts now represents a fourfold increase in non-malicious insider actions detected across this year’s dataset of more than 22,000 breaches globally.”
AI skepticism grows among US youth - Along with datacenter backlash, 2026 and 2027 is shaping up to have a big, general population backlash against AI. “Polls show that 70% of Americans think AI is moving too fast, over 50% have negative views of it, and just 18% of young people say they feel hopeful about it. Partly, they are turned off by AI’s upending of the job market. ‘Every other day, a new AI agent is being released in the market,’ said Vaishali Hireraddi, 23, a University of California, Davis, graduate student who’s applied to 500 jobs so far. ‘What am I doing with my life?'”
Your Slop, My Sludge - AI sludge: the build up of comprehension debt and day two toil due to AI generated apps.
The coming AI backlash
Rampant shadow AI
🤖 Against the Tyranny of the Apollonian - Beautiful becomes meaningless when machine-generated beauty is infinite; the human response is an embrace of messiness, imperfection, and visceral creativity as a cultural immune reaction. // Summary
The VW ID Buzz: six months and eight thousand miles later - “it feels like you’re driving around in a giant phone booth, but in a good way” Seems like a great car: “We don’t regret anything about this purchase and I bet we’ll have this car for the next ten years”

From: AI skepticism grows among US youth

AI Summaries

Things I asked the robot to summarize for me.

From: Economic history as it's happening is alway relative

Wastebook

“I can think of few better ways of raising social welfare than making sex 10% better!” Repugnant Economics
“Good marketing is putting true facts in a way people want to hear.” John Gruber on Dithering, May 19th, 2026.
“This is a three-beer conversation mistaken for a finished essay.” Observations on Writing with AI
“reflation” China’s narrow reflation

ICYMI

Use AI to write more shorter pieces

Logoff

Perhaps you have noticed something new above: links to AI summaries. As I mentioned in my “please use AI to write shorter shit” post above, I have the robot summarize a lot of things. I have been debating whether or not to publish those. I mean, I love publishing - just publishing anything. So, I see that big stack of content and it pains me to not publish it. But people hate the fuck out of AI.

So, I have done the following. I started selectively publish summaries on incomprehensiblemedia.com. In this newsletter, I’ll include links to summaries for the links I have above, plus a new section that lists the some of the other summaries I’ve run.

This is a blog with an RSS feed. I set the robots.txt to exclude crawling of all types. I’m intending to capture traffic from the original. So while it’s not “hidden,” hopefully it will not damage the OP’s Google juice - or LLM juice…or whatever juice the authors like.

Ideally, I would automate this even more so that I can read the summaries in the same flow as reading my RSS feeds. Also, I’m occasionally generating podcasts of summaries. These have a different script written to be more natural voice and use an AI version of my voice - my thinking on that last one is that way I won’t piss anyone off but myself.

Eventually, I need to publish these.

My hesitation for posting this stuff - beyond the juice concerns above - is the AI posting something stupid or, worse, shit-talking about my work, etc. It’s very annoying to have that bottleneck what would otherwise be a nice flow of reading: stuff just shows up in my RSS feeds and podcast client.

All that AI governance stuff is real!

Anyhow, there it is.

And, finally:

From: Hotel Retro: Vintage Luggage Labels from Tokyo to Buenos Aires

Want to subscribe to this newsletter and get it in your email? Do that here. You’ll just get this type of link and post round-up, not everything posted on the weblog.

The coming AI backlash

Wed, 20 May 2026 07:15:53 +0200

Along with datacenter backlash, 2026 and 2027 is shaping up to have a big, general population backlash against AI.

Polls show that 70% of Americans think AI is moving too fast, over 50% have negative views of it, and just 18% of young people say they feel hopeful about it. Partly, they are turned off by AI’s upending of the job market. ‘Every other day, a new AI agent is being released in the market,’ said Vaishali Hireraddi, 23, a University of California, Davis, graduate student who’s applied to 500 jobs so far. ‘What am I doing with my life?’

🔗 AI skepticism grows among US youth

Rampant shadow AI

Wed, 20 May 2026 06:47:41 +0200

Of the 45 percent of all professionals using AI in the workplace regularly, 67 percent of those were accessing the platforms using personal accounts that were not authorized by their IT teams, data from Verizon’s annual data breach investigations report showed.

Verizon said that the proportion of users accessing AI through personal accounts now represents a fourfold increase in non-malicious insider actions detected across this year’s dataset of more than 22,000 breaches globally.

Better get an AI platform in place.

🔗 Shadow AI surges in the workplace

Use AI to write more shorter pieces

Tue, 19 May 2026 07:36:38 +0200

While people are all excited shout using LLMs to write - which is awesome, I am all about AI empowering people to do things they didn’t have the agency to do, writing being one of them - I am imploring y’all to get the LLMs to write more concisely. I am looking at all y’all Substack writers with my sternest Strunk & White spectacles.

The number one use case I have for AI, used countless times a day, is “wow, my digital brochacho, this is really long and convoluted. Can you rewrite this to be more to the point and shorter? What are they trying to say?” There’s a lot of content out there that I want to read because I can smell that there’s something interesting in it, but it comes wrapped in a thick layer of concrete that I have to chip away at.

It works amazing well, but is an odd “your AI is talking to my AI” step.

And LinkedIn hustlers. Come on now. Just tell the LLM to not write like a LinkedIn post and cut it with the staccato one line paragraph thing. Please stop tommy gunning me with your life and career advice. You’re allowed a one line paragraph per 1,200 words. No need to be LinkedIn Hemingway.

Tokenmaxxing at Amazon, Potato Stamp Fonts, and Tax Code Hacking - Related to your interests, Monday

Mon, 18 May 2026 09:56:49 +0200

Also: Elon’s OpenAI lawsuit, build-vs-buy for agentic AI, and rethinking observability.

From: Wakamai Fondue, the tool that answers the question what can my font do?

Amazon employees admit to using AI unnecessarily to pump up internal usage scores - workers complain of intense pressure to use AI tools - Tokens are the new lines of code. Also, yes, of course, “Goodhart’s Law," blah blah.
The hidden cost of build vs. buy for agentic AI in regulated industries - “Building an internal agentic AI platform in banking or insurance demands a multi-year orchestration engineering commitment with a regulatory surface area that most organizations underestimate”
Why AI is forcing enterprises to rethink observability
How Dangerous Is Anthropic’s Mythos AI? - Using AI to hack tax codes is a bigger threat than hacking computers.
Elon’s OpenAI Lawsuit Is Boring and Insulting, and It’s Already a Success - Winning in the court of opinion, and all that: “Musk, to be clear, has already succeeded. Regardless of what actually happened and what’s decided from here, he’s inflicted real pain. The trial has been an opportunity to make Altman and OpenAI’s founders look like greedy, sociopathic liars, which is a narrative that much of the public wants to believe.” // And, it’s worth asking if things could have turned out any other way than a non-profit kind of contorting its way to becoming a for-profit: “So go back to 2017. Would the world be better off if OpenAI had remained a nonprofit and slowly withered on the vine while being outspent 100 to 1, as Google secured a monopoly in the most powerful technology of the 21st century? And in that scenario, how much longer does it take for us to get great AI products if Google has no competitive pressure to gets it act together?”
How to raise money without profit, and then profit
Tax code hacking - maybe the biggest AI threat is tax scamming.
Tatermaxxing - “We had the idea to make a Bodoni interpretation with potato stamps, so we bought 8kg of potatoes, some knifes and carved a long, long evening in the kitchen. When we finally had the full alphabet we stamped it on paper, made a font out of this and called it Bodedo.” // Maybe my new mindset on presentation should be: an excuse to use a cool font.

From: At MoMA, Duchamp Vanishes Into the Shadow of His Own Legend

Wastebook

“bonkbuster” Here
“But what if I buttchug one potion and drink the other?” Here
“The bus stopped, and Bill Gates stepped away from the steering wheel. It wasn’t a bus at all, but a yacht. I’d been pranked. This was my graduation from the grind. I’d made it. He gave me the yacht as thanks. The old lady was actually Jeff Bezos, he was in on the prank. He acquired my business immediately and now I’m sipping on champagne bottles. All thanks to Claude.” LinkedIn fanfiction
“Whether or not that is true, Detroit persuades me that other places should engage in more plaquemaxxing.” Here
“Carousel of justifications” Here
“Future Asteroid farmer.”

ICYMI

Got your deck ready? - my thoughts of slides as the primary corporate collaboration tool.
Buy your platform, AI edition
5G versus The Grapes of Wrath
Aesthetic Computing
Datacenter NIMBYism: What Did You Think Was Going to Happen?
Scenes from the weekend
The world isn’t curl - Software Defined Talk #572 - “This week, we discuss how security gets sold to execs, where agentic coding and security collide, and Cloudflare vs. Datadog’s diverging paths. Plus, Coté weighs in on sugar cookies.”

Logoff

There is a lot of “what do I want?” thinking going on in my head recently. It is exhausting, and I’m not really interested. But Know that if I don’t resolve it, I’ll both regret it and resent past-me for not dealing with it.

Also, I hate calling people on the phone. There are some phone calls that would take just 10 or 30 minutes on my to do list that have literally been there for years. My life would be so much better if I could just pick up the phone - or “tap” it, I guess - and say, “Hello there, I’ve got this thing I need help with…”

Want to subscribe to this newsletter and get it in your email? Do that here. You’ll just get this type of link and post round-up, not everything posted on the weblog.

Mon, 18 May 2026 09:54:46 +0200

Weekend of May 16th, 2026 to May 17th, 2026.

Tatermaxxing

Mon, 18 May 2026 07:19:29 +0200

We had the idea to make a Bodoni interpretation with potato stamps, so we bought 8kg of potatoes, some knifes and carved a long, long evening in the kitchen. When we finally had the full aphabet we stamped it on paper, made a font out of this and called it Bodedo.

Maybe my new mindset on presentation should be: an excuse to use a cool font.

🔗 HVD Bodedo

Buy your platform, AI edition

Mon, 18 May 2026 07:13:51 +0200

Building an internal agentic AI platform in banking or insurance demands a multi-year orchestration engineering commitment with a regulatory surface area that most organizations underestimate. [Bryan Ross]

Tinkers and opexmaxxers take in huge risks when they decide to build their own platforms. And it usually fails, for at least seven reasons.

🔗 The hidden cost of build vs. buy for agentic AI in regulated industries

5G versus _The Grapes or Wrath_

Mon, 18 May 2026 06:42:55 +0200

From the recent cellphone ban in schools study:

The share of students using cellphones in class for nonacademic reasons declined to 13 percent from 61 percent in schools using the pouches, according to teacher surveys, which suggested that students were not widely able to circumvent the bans.

Even so, the bans had a “close to zero” effect on test scores, according to the paper. Test scores are affected by many factors, including the stability of students’ home lives and the quality of teaching and curriculum. The researchers also noted that once cellphones had been banned, students might have been distracted by other forms of technology, such as laptops, which are ubiquitous in American classrooms.

The bans also did not improve student attendance or perceptions of online bullying. And in the first year after strict bans went into place, student suspensions increased by an average of 16 percent — a large and troubling change.

This study was also covered on The Political Gabfest a few weeks back. As Emily points out, at least the teachers are happy:

One positive benefit is that teachers are more satisfied, the teachers are happier, they have to spend less time policing it. That seems like a genuinely important issue. You want teachers to be happy, they will stay longer, they will do their job better, they will, you know, be able to teach more because they’re not policing, they’re not copying the kids who are sneaking phone use. That’s really important. The other one was, did you guys see that, that wonderful little detail? That there was one school district that looked at school library use in the wake, in the wake of the cell phone bans and that way, way, way more books were checked out of school libraries after phone bans, which is fascinating. That seems like a pure good.

Studying ongoing “adult” use would be more interesting. What the relationship between cellphone use and income, happiness, etc.

Also, I don’t think “cellphone” is the right framing, but what tot do with it. Surely there’s a difference between reading Facebook and Twitter al day versus Wikipedia and Chaucer in the a kindle app.