I think that’s wrong on both counts. Amazon’s slowdown is mostly a calendar quirk. And whatever AI is doing to these ad businesses, it isn’t coming from Muse or ChatGPT shopping on your behalf. It’s coming from inside the house.

Amazon’s ad business is fine

Amazon moved Prime Day from July to June this year. That shifted the biggest ad week of the year from Q3 into Q2, which is why Q2 grew 26% and Q3 grew 11%. Comparing either quarter alone to last year tells you nothing.

Add them together and Amazon sold $39.4 billion of ads in Q2 and Q3, up 18% from a year ago. The same two quarters in 2025 grew about 23%.

Is that a slowdown? Yes, and I’m not going to pretend otherwise. The Q4 guide is “high teens” against 22% last year, partly because Amazon is now lapping its first NBA season and last year’s sponsored search changes. But high-teens growth on a $68 billion ad business is not a business being killed by anything.

Outside agents barely show up in the data

Here’s what the companies actually said:

• Amazon: Andy Jassy said outside shopping agents aren’t a factor because “few customers are doing much shopping this way yet, as the experience is still not great.”
• DoorDash: On the call announcing the Instacart deal, management said third-party AI agents have produced several hundred orders. Total. Since launch.
• Amazon, again: It blocked Muse, is suing Perplexity, and has moved against OpenAI’s and Google’s agents. It’s hard for an agent to steal ad impressions from a store it can’t get into.

Yes, Amazon and DoorDash would both rather you keep opening their apps. Discount the quotes. They still line up with each other, and with every order log I’ve seen.

The real disruption is the agents they own

Now look at Google. Philipp Schindler said AI Mode query growth hurt search revenue “because of the lower monetization we were seeing in AI Mode.” Meanwhile, queries grew at their fastest rate in five years and paid clicks grew 8%.

Read that again. People are searching more than ever. More of that searching happens on a Google surface with fewer ads.

Amazon has the same dynamic, it just doesn’t say so. Alexa for Shopping had more than 400 million users in the last 12 months, and interactions grew 570%. Every shopper chatting with Alexa for Shopping is a shopper who isn’t scrolling past a wall of sponsored listings. Amazon doesn’t disclose that tradeoff, so this part is my inference. But it’s a far better suspect than a horizontal agent that’s sent DoorDash a few hundred orders.

So if agentic AI is denting Google and Amazon ads, it’s their own vertical agents doing it: AI Overviews, AI Mode and Gemini at Google, Alexa for Shopping at Amazon. That’s a very different problem from “the agents are coming.”

And it’s a problem they’ve solved before

Google waited two years to sell AdWords. Google Shopping was free until 2012. Amazon took 17 years to roll out Sponsored Products. Facebook told IPO investors in 2012 that it barely made money on mobile, and mobile was half its ad revenue by the end of 2013.

The playbook never changes. Give customers a lot of value first. Build the habit. Then add ads as people come to depend on the service and tolerate more of them. Both companies are already running it: Amazon gave away sponsored placements in its assistant before it started charging, and Google is scaling ads in AI Mode and AI Overviews right now.

Judging the long-term ad value of AI Mode by its first year is like judging mobile by Facebook’s 2012 S-1.

The data behind the panic is thin

A lot of the “agents are taking over” evidence doesn’t hold up when you open the hood.

• McKinsey’s new survey says 17% of U.S. AI users bought something inside an AI tool. The U.S. sample is about 1,000 people, “AI tool” includes Amazon’s own assistant, and it’s what people say they did, not what they did. Stated preference is not observed behavior.
• Adobe says AI traffic to retail sites is up 393%. Off a tiny base, from a company that sells AI optimization tools.
• HUMAN Security says retail gets more agentic traffic than any other industry. HUMAN sells bot management.
• Profound estimated Muse traffic from unverified bot requests and calls its own method a proxy. Profound sells AI visibility tools.

I’m not saying any of them is lying. I’m saying the methods are black boxes and every incentive points the same way. When the survey and DoorDash’s order log disagree, I’m going with the order log.

Someone still has to pay for the agents

Every one of these services needs a business model: subscriptions, ads, or a cut of each sale. We’ll see all three, mostly as hybrids. But I’m with Ben Thompson that ads are the model that wins at consumer scale. Retail margins can’t absorb another take rate, and most consumers won’t pay for software. “Ads” here just means a brand paying the agent’s owner for better placement: a sponsored prompt, an offer, a spot on the shortlist.

Thompson also makes the best case against me. He argues agents, not chatbots, are the real threat to Google and Amazon, because once the human leaves the loop, nobody sees the ads. In the long run he might be right about how ads change shape. But today the human is still in the loop almost everywhere, and the ad dollars will follow the decision wherever it goes.

What I might have wrong

If Amazon’s Q4 comes in well below high teens with no timing excuse, or if Google keeps adding ads to AI Mode and the gap between query growth and paid clicks doesn’t close, then the slowdown is structural and I’m too optimistic.

What to do first

• Don’t cut Google or Amazon search budgets because of the agent narrative.
• Test ads in AI Mode, AI Overviews and Alexa for Shopping while the auctions are thin. Early Sponsored Products buyers got cheap clicks for years.
• Treat outside agents as a product data project, not a media line, for now.
• Ask every agentic research vendor for its sample and methodology before it goes in a budget deck.

So here’s my question: has anyone actually seen measurable ad impressions or sales lost to a third-party agent? Or are we all just reading the same vendor charts?

Sources

• Amazon Q3 2026 ad revenue and Prime Day timing, PPC Land
• Amazon’s ad revenue growth decelerates, Marketing Dive
• Jassy on outside shopping agents and Alexa for Shopping, The Next Web
• Google’s AI overhaul of search hits snags as ad growth cools, Marketing Dive
• Alphabet Q3 2026: search ad growth slowed, queries did not
• Amazon is giving away Rufus ads for free (for now), The Drum
• Google Shopping’s transition to pay-for-play, TechCrunch (2012)
• Facebook now makes most of its money from mobile ads, Business Insider (2014)
• Who wins when AI does the shopping?, McKinsey

So here are my receipts. Literally.

My own order history

I placed my first Amazon order in 1995. It was the only one I placed that year. For most of the next decade I averaged a few orders a year, and in 2002 I didn’t place any at all.

Then the curve bent:

• 2009 was the first year I averaged about one order a week (51 orders).
• 2015, the year I coined the phrase, I placed 171 orders. That’s one every 2.1 days.
• 2020 more than doubled 2019, from 170 orders to 389.
• In 2026, through November 3, I’ve placed 444 orders. That’s one every 17 hours, and a pace of roughly 525 for the year.

That’s one shopper, and it’s Amazon only. It also understates my household’s real behavior, because it doesn’t count the orders other people in my house place, or anything bought from Target, DoorDash or a brand’s own site. And we’re a Walmart+ family (now with a Sam’s Club membership, too), so this chart doesn’t even capture the majority of our groceries.

I know a single person’s order history is an anecdote, not a dataset. But it’s observed behavior, not a survey answer. Nobody asked me how often I planned to shop. The order log just recorded what I did.

“No one’s ever going to buy that online”

Over 30 years I’ve sat in a lot of boardrooms where smart people explained why their category was safe. The list kept changing. The confidence never did.

• “No one will ever buy videos online.” Digital accounted for 92.4% of the $62.2 billion Americans spent on home entertainment in 2025, while disc sales fell to $870 million, according to the Digital Entertainment Group. (I launched e-commerce at Blockbuster, so I take this one personally.)
• “Fine, but no one will buy a TV online.” Online sales were 39.0% of Best Buy’s domestic revenue in the fourth quarter of fiscal 2026, according to the company’s earnings release. That’s the retailer with a big box in nearly every market, and it still doesn’t count the TVs sold on Amazon and Walmart.com.
• “Electronics, sure. But no one buys clothes online.” E-commerce will account for 37.9% of U.S. apparel, footwear and accessories sales in 2026, according to EMARKETER.
• “Jason, we sell beautiful things for your kitchen table. No one buys that online.” E-commerce was more than 65% of Williams-Sonoma’s revenue in fiscal 2025, according to CFO Jeff Howie on the company’s fourth-quarter call.
• “Who would ever buy an engagement diamond online?” Blue Nile built one of the biggest engagement ring businesses in the country by selling mostly online, which is a big part of why Signet, the largest mall jeweler, bought it in 2022.
• “No one’s ever going to buy a car online.” Tesla’s sales channels “currently include our website and a global network of company-owned stores,” and it has no independent dealer network, according to its 2025 10-K.

The fair counterpoint is that a lot of these sales still involve a store. People still (sometimes) test drive the Tesla. They look at the ring under the showroom lights. That’s true, and it’s why I care more about digitally influenced sales than about where the transaction lands. But the “no one will ever” crowd wasn’t arguing about channel mix. They were arguing that consumers would never change. Every one of them was wrong.

By now, anytime I hear “no one’s ever going to,” my spidey sense starts tingling.

Why my purchase behaviors changed

Nothing about my personality changed between 1995 and 2026. The friction did:

• Ordering got easier. When I started in e-commerce in the ’90s, the average online purchase took about 28 clicks. Now it takes one.
• Assortment got deeper. The endless aisle actually became endless.
• Prices got easier to compare, so the trip stopped being the way to get a good deal.
• Speed of delivery went from a week to a day to a few hours.

I still go to stores (and no one loves stores more than me). But I used to solve most problems by driving to the mall, the grocery store or the club store, and I batched my needs to make the trip worth it. Now, more often than not, I order what I need, when I need it, from whatever screen is closest. That’s what “always shopping” means. It isn’t more shopping. It’s shopping that has dissolved into the rest of the day.

Grocery is next

Grocery was supposed to be the exception. Perishables are hard, baskets are big, margins are thin, and people like to pick their own avocados. All of that is still true. But the speed problem is getting solved.

Walmart says it can reach about 95% of U.S. households in under three hours, up from 76% two years earlier, according to its third-quarter fiscal 2026 results. Its U.S. e-commerce sales grew 24% in the second quarter of fiscal 2027, and store-fulfilled delivery grew 40%, according to the company. And the business is now profitable, which was the thing skeptics said would never happen.

Amazon expanded same-day delivery of perishable groceries to more than 2,300 U.S. cities and towns by the end of 2025, and said its online perishable sales grew 30 times over that year. In 2026 it rolled out Amazon Now, 30-minute delivery of groceries and essentials from small neighborhood hubs, to dozens of U.S. cities. (Amazon also says customers who add fresh groceries shop about twice as often. That’s a correlation. The people who buy fresh online were probably already its most frequent shoppers.)

The money is moving, too. Online grocery sales grew more than 20% in the first half of 2026 while sales at food and beverage stores and general merchandise stores combined grew less than 2%, according to Brick Meets Click.

The casualty is the weekly stock-up trip. Nobody needs to predict the 72 items they’ll consume over the next seven days and haul them home in one exhausting trip when anything they run out of can show up in 30 minutes to four hours. We don’t go grocery shopping anymore. We are always grocery shopping.

That changes the economics for everyone built around the trip. Smaller, more frequent baskets mean fewer impulse buys at the endcap, fewer chances for a new brand to get discovered on the shelf, and a lot more value in being the default reorder.

The next “no one’s ever going to”

Here’s the one I hear constantly now: “Jason, no one’s ever going to trust the robot to make purchases for them.”

I’ll give the skeptics their best argument. Delegating a decision is different from changing a channel. Every shift above still had a human picking the product. Handing that choice to an agent means trusting it with your money and your taste, and a bad substitution on Tuesday’s dinner is more memorable than a hundred good ones. I also think agentic commerce will take longer than the hype suggests.

But we already delegate more than we admit. The robot already picks the music we listen to. Subscribe & Save picks the date. The grocery app picks the substitute. The default reorder button picks the brand. Delegation won’t start with engagement rings. It will start with paper towels, dish soap and the 30 items most households buy on repeat, exactly the items the weekly trip used to be built around.

Thirty years of “no one’s ever going to” makes me suspect this one ends the same way.

What to do first

1. Sort your assortment into replenishment items and considered purchases. The replenishment items are the ones that move to always-on ordering first, and to agents after that.
2. Stop planning around the trip. Measure purchase frequency and how often you’re the default reorder, not just basket size and store traffic.

What’s the “no one’s ever going to” you’re hearing in your boardroom right now?

Fine. Now take out gas stations, then take out price, and here’s what’s left: the store-based half of American retail grew about half a percent this year. That’s your TLDR;

August against last August

Total retail trade grew 5.5% over August 2025 in dollars, and 1.5% after price. Core retail, which strips out gas stations and auto dealers, grew 4.9% in dollars and 2.0% in real terms.

The gap between those two toplines is gasoline. Gas station sales were up 20.5% in dollars and down 5.4% in real terms. Americans bought less fuel in August than a year ago and paid a lot more for it. Roughly a third of the total retail dollar growth in August walked through a gas station.

A few other August reads worth pulling out:

• Nonstore retail: +10.4% in dollars, +6.2% real. Still growing at twice the pace of core retail.
• Grocery: dollars fell 0.5% against last August. With food-at-home prices up 2.2% for the year, that’s a 2.7% real decline, the worst month of a year in which grocery hasn’t had a single positive real month.
• Apparel, in the back-to-school month: +3.5% in dollars, -0.1% real. All price. Apparel’s real growth has faded every quarter this year, and August is the first month it went negative.
• Home improvement: +5.2% in dollars, +3.8% real. That’s three straight months of real growth after five straight months of decline.
• Electronics: +7.9% in dollars, +8.5% real, because electronics prices keep falling.

January through August

Eight months in, the year-to-date story hasn’t moved much. Core retail is $3.55 trillion, up 5.2% in dollars and 2.2% in real terms. Total retail trade is $5.17 trillion, up 5.3% and 1.9%. About 57% of this year’s core retail growth is price.

The range I gave you last month for inflation was wrong. My assumption was that All Items overstated inflation for home centers and general merchandise because it’s a third shelter. It doesn’t. Tools and hardware inflation is running hotter than the economy, so I’ve started adjusting for inflation on a segment by segment basis. If you do your own deflation adjustments, pick your deflators carefully. I’d rather tell you that than quietly move the number.

The two retail economies are channels now

I’ve been describing “two retail economies” for a while, meaning categories growing in real terms next to categories shrinking. Eight months into 2026, the cleaner split is by channel.

• Nonstore retail: $1.06 trillion year to date, up 10.3% in dollars and 6.6% in real terms. It’s now 30% of core retail, up 1.4 percentage points in a year.
• Nonstore accounts for 56% of all the dollar growth in core retail this year. $99 billion of $175 billion.
• Everything else in core retail, meaning every store-based category combined: up 3.1% in dollars and 0.5% in real terms.

Half a percent. That’s the store-based economy, in aggregate, after price. And that figure is generous, because store-based retailers book their own e-commerce inside their store categories. Big box e-commerce sales land in general merchandise, not nonstore. Strip those out and physical stores are almost certainly negative in real terms this year.

The usual caveat: nonstore is a proxy for e-commerce, not a measurement of it. It includes vending, fuel dealers and direct selling, and misses the digital sales of store-based retailers. The cleaner number is the quarterly e-commerce report, which had second-quarter e-commerce up 12.4% and at 16.4% of total retail, up 0.8 percentage points, according to the Census Bureau. Same story, different instrument.

The category picture

Nominal and real, January through August:

• Gas stations: +15.7% / -3.0%
• Sporting goods and hobby: +10.1% / +7.2%
• Online and mail order: +10.3% / +6.6%
• Electronics and appliances: +6.7% / +7.4%
• Apparel: +5.4% / +1.7%
• Home improvement: +4.6% / +0.8%
• Restaurants and bars: +4.2% / +0.6%
• General merchandise: +3.1% / -0.4%
• Health and beauty: +1.9% / +1.2%
• Auto and parts: +1.9% / +2.6%
• Grocery: +0.7% / -1.7%
• Furniture and home: -1.6% / -3.9%

Four categories are shrinking in real terms. Furniture down 3.9% which is the worst category in retail and still shows no sign of turning. Home improvement, showing some strength. When people can’t move, they fix.

General merchandise at -0.4% real is the one I’d watch most. This is Walmart, Target, Costco and the dollar stores, combined, running slightly negative after price on a generic goods deflator. It’s a coin flip around zero, not a confirmed contraction. But the largest store-based category in America is not growing in real terms, and it hasn’t been in five of the last seven months.

Electronics is the one category where the real number beats the nominal number. Prices fall, so the same dollars buy more. Some of that is more devices and some of it is BLS quality-adjusting a better device at the same price. Don’t call it unit growth. But it’s the healthiest looking category in the data, and it has been every month this year.

What to do Monday morning

Split your growth three ways: price, channel, and everything else. If you’re a store-based retailer running a 3% comp, you’re at the category average, and the category average is roughly zero after price. That’s not a criticism. It’s the baseline you should be measuring against, instead of the 5% topline in the press release.

Check your gas exposure in the consumer wallet, not just your P&L. Americans have spent $65 billion more at gas stations this year than last for fewer gallons. That money came out of something, and the grocery and general merchandise numbers suggest where.

The question I keep coming back to: if nonstore is 56% of the growth and stores are flat after price, is a 5% retail year actually a good year for anyone who isn’t a marketplace?

I built these two images because I wanted to actually look at the thing we keep calling fragmentation. And I think we describe it wrong. Content was already fragmenting in 2006. Cable, TiVo, early YouTube. The audience was splitting well before the iPhone existed.

What broke is co-viewing. Four people in one room used to be one audience. Now it is four.

Before anyone tells me mass reach is dead, it isn’t. The Super Bowl still does exactly what it always did. So does a finale people decide to watch at the same time. Ehrenberg-Bass has been right about this for forty years: growth comes from reaching people who don’t buy you yet, and that takes reach.

Here is the actual change. We used to inherit a shared culture. Now we have to build one.

Inheriting was free. Buy the right dayparts and you were in that living room by default. Building is not free. It means paying up for the few moments that still assemble people, and then doing the unglamorous work of showing up in four separate feeds for everybody else.

Most media plans I look at are still written as though the first panel is true.

Which panel is your 2027 plan built for?

I got to that number the hard way, because nobody publishes it. I wanted to answer what sounds like a simple question. How many Americans use AI chatbots to find and buy products?

The published answers disagree by a factor of ten.

The estimates run from 2% to over 20%

Here is the spread.

• 2.1% of messages are about a product someone can buy, according to the paper OpenAI’s own researchers published with academic economists in September 2025.
• 5.2% of conversations are about buying things, according to Google’s ATLAS report from August 2026, which covered 15 million conversations with Gemini.
• 21.6% of sessions show some commercial interest, according to a research panel called Measure Protocol. Only 7.1% showed a real signal that someone was close to buying.
• 19.2% of conversations are commercial, according to Profound, a company that sells AI visibility software.
• “Roughly 20% of queries in ChatGPT have a direct commercial intent,” according to OpenAI’s head of global ad sales, speaking at Cannes Lions in June 2026.

Four of those five are about ChatGPT. Same rough time period. Numbers ten times apart.

Two definitions are doing all the work

The first problem is what counts as shopping.

“What laptop should I buy for under a thousand dollars” is clearly shopping. What about “help me plan a beach trip”? What about “how do I fix a leaky faucet”? The beach trip ends in a hotel booking. The faucet ends in a hardware store run. OpenAI’s researchers count only the laptop, and that gets you 2.1%. Count anything that touches a buying decision and you get 20%.

Neither one is wrong. They are just answering different questions.

The second definition is what you are counting. A message is the actual thing you type. A conversation is the whole back and forth, usually about five messages. Profound counts conversations. OpenAI’s researchers count messages. Mix those up and your answer moves by five times, in whichever direction happens to flatter your argument.

Watch who is doing the counting

Two of these numbers come from scientists. The OpenAI paper and Google’s ATLAS report both publish their sample sizes, their methods, and the exact instructions they gave their classifiers. You can check their work.

Most of the others come from companies that do better when the number is big. That does not make them wrong. But it does mean you should want to see the method, and in most cases it has not been published.

The 20% figure needs the most care. It was said out loud at an advertising festival by the person who sells ChatGPT’s ads. There is no paper behind it. No sample, no time period, no definition of what counts. It also happens to be the highest defensible number available.

Here is the part that bothers me most. OpenAI has not said how many messages ChatGPT handles since July 2025. That figure is fourteen months old now, which is about twenty years in AI time. There has never been a US-only version of it. The one number you most need to size any of this does not exist.

My best estimate, and how I got there

Start with 2.5 billion ChatGPT messages a day worldwide, which is OpenAI’s last published figure. About 19% of ChatGPT’s traffic is American. Group those messages into conversations. Then apply the commercial share.

You land at about 566 million commerce conversations a month in the US, with ChatGPT alone.

I checked that a second way. Profound publishes its own worldwide count of commercial conversations. Take the US share of that and you get 448 million. So the range I use is 448 million to 566 million.

Remember that this is one chatbot. Gemini, Copilot, Claude and Perplexity are not in the number at all. Comscore has ChatGPT at roughly 69% of American AI assistant users, so counting everybody would push this up by a third or more.

Some of my assumptions can’t be validated (yet):

The first is that a conversation runs about five messages. That ratio is implied by the OpenAI research paper rather than stated in it, and if the real number is four or six, my estimate moves by 25% in either direction. The second is Profound’s 19.2%. Profound has never published the rules it uses to decide what counts as a commercial conversation, so I am trusting a classification I cannot inspect, from a company that sells software to brands who want to show up in AI answers. I used their number because a second, independent route landed close to it. But that is reassurance, not proof.

This should get better. Similarweb and NIQ say they will start measuring shopping inside ChatGPT and Gemini by the end of this year. OpenAI could publish a current message count tomorrow and settle half of this. Until somebody does, everyone working in this space is estimating, and the honest move is to show the arithmetic so we can all argue with it.

What this looks like compared to e-commerce sites

Monthly US visits, according to Similarweb, with my ChatGPT estimate slotted in at the middle of its range:

• amazon.com 2,300 million • ebay.com 512 million • ChatGPT shopping conversations 448 to 566 million • walmart.com 492 million • temu.com 357 million • etsy.com 250 million • target.com 226 million

ChatGPT sits right alongside eBay and Walmart, and at about a quarter of Amazon. That is a real number. It is also not the number some people want it to be.

It is fair to ask whether any of that turns into a sale. ChatGPT can take your money now. Instant checkout is live, and Best Buy checkout showed up in the wild this week. But OpenAI has never published a transaction count, only about 30 Shopify merchants were actually live in February against a launch promise of more than a million. For now this is a discovery channel with a cash register in the corner.

So this is not ChatGPT eating retail. This is ChatGPT showing up as a large commerce property, about the size of the second tier.

Where I could be wrong

A conversation is not a visit. Someone on walmart.com has already decided to shop. Someone asking ChatGPT about running shoes may be a long way from buying anything. The comparison is fair on units. It is not fair on intent.

The traffic AI actually sends to retail sites is also still tiny. Amazon got about 0.16% of its visits from AI referrals in June 2025, according to Similarweb. Adobe reports that AI referred traffic to US retail sites more than doubled in the year to May 2026, but has never published what share of total traffic that represents. Fast growth on a base nobody will state.

One thing cuts the other way. This is ChatGPT alone, so every other assistant is upside on the number.

What to do about it Monday

Stop asking whether customers are shopping inside AI. Enough of them are that it is worth your time now.

Start asking whether your products show up when they do. Your product data, your reviews and your specs have to be readable by a machine that will boil them down to two sentences. Most product pages are still written for a shopper with a browser and time to scroll.

Anyone quoting one of these numbers at you should be able to tell you which one they mean, and where it came from.

Which number have you been using?

I’ve had a personal interest in productivity since I first found David Allen’s Getting Things Done in 2001. I spend most of my working hours on a Mac and, like most of us, I use a wide variety of tools to make the work more comfortable. But there’s a small handful that give me actual joy every time I use them. Mostly because they take the friction out of something I do constantly, or because they eliminate something that used to be reliably annoying. The best of them solve a problem before I’ve consciously noticed it was a problem.

This isn’t my usual content. I’m posting it mostly because I want the reverse: I want to hear what’s on your list that isn’t on mine.

One disclosure before I get into it. The Raycast link below is an affiliate link, and it’s the only one in this piece. I’m using it because it gets you 10% off if you decide to subscribe, which seemed worth passing along. It also pays me a commission, and I’d rather say that up front than have you wonder. Every other link here is an ordinary link that earns me nothing.

Raycast

*[IMAGE: Productivity_Tools_Logos/raycast.png • link: https://raycast.com/?via=jason-goldberg (affiliate, 10% reader discount)]*

I’ll start here because it’s the tool I probably use most in a normal day.

Raycast is in the category people call Spotlight replacements. You hit Command-Space, you get something much better than Apple’s search, and then it keeps going well past search.

The first tool I used in this category was Alfred, and I was an avid user for years. Alfred is still genuinely good. I think a lot of us drifted to Raycast because we got impatient with the pace of modernization, and I’ve never looked back.

Raycast does far too much to list, and I suspect every user has their own shortlist. Mine:

• It’s my app launcher. Command-Space, a few letters, I’m in the app.

• It’s my file search. A few words and it finds the file.

• It’s my hotkey manager, which means I rarely type an app name at all. Mac modifier keys get messy fast once you’re stacking Command, Option, and Control. Raycast lets you set up a hyper key, and mine is Caps Lock. Tapped once, Caps Lock still works normally. Held down, it’s a modifier. Hyper+F is Finder. Hyper+Z is my browser. That one change reorganized how I move around my machine.

• It’s my window manager. Hyper plus an arrow key throws the focused app to a half, a quarter, or another monitor. I reconfigure two screens all day long.

• It’s my snippet library and my clipboard history, so anything I’ve copied recently is searchable.

• It’s now the front end for most of my quick AI queries. I used to run those in Gemini. Now I type the prompt into Raycast, hit Tab instead of Enter, and it goes to whichever model I’ve chosen. I can share the screen I’m looking at, which I do constantly when troubleshooting. And it’s easy to build small skills: highlight an email, have it fix the typos and grammar in my voice, paste the clean version back.

The extension ecosystem is the other half of the value. Raycast talks to most of the other tools I use. My favorite is an extension called 2FA. Apple already does a nice thing where a texted authentication code gets offered to you automatically in the browser. It doesn’t work for emailed codes. The 2FA extension checks my Gmail and Outlook for recent authentication emails, pulls out the code, and offers to type it. Every time I don’t have to tab to an email client, find a code, copy it, tab back, and paste it, I say a small thank you.

And I think Raycast is a model of how to run an app company. The CEO and the developers are visibly active. They publish real teaching content. They run a Slack channel and answer in it. It feels like they’re actually trying to get ahead of what their users need.

CleanShot

My favorite screen capture tool, and I take an enormous number of screenshots for work.

It’s organized in a way that fits how I work. Captures go into my native file system and onto the clipboard by default. I have separate hotkeys for the different capture types I use often, including full scrolling website captures, which are genuinely useful in my line of work. The annotation tools are good and I use them daily. The video capture and editing features have gotten interesting.

I used Snagit for many years and it’s a good product. It also got quite bloated. CleanShot has all the utility I actually used, runs much faster, and integrates cleanly with the Mac and with Raycast.

1Password

I haven’t typed a password in a very long time.

I’ve used password managers for years and I’ve moved my whole family onto one, including the older adults in my life I help support, which has been a relief for everyone. 1Password is my current choice.

When someone asks me what my password is for something, the honest answer is that I don’t know. 1Password generates a complex one for every service and types it for me. I’ve never seen most of them.

It’s also my authenticator. All the six-digit codes moved there, which means they’re available on every device I own instead of trapped on one phone, and they’re backed up. And it holds my passkeys, so more and more logins are now a single quick approval. It’s the first utility I install on any new machine, because without those credentials I can’t do anything.

PopClip

A small utility I’m not sure enough people know about, and one I use nearly as often as Raycast.

It hooks into the native Mac highlight behavior. Whenever you select text, PopClip works out what kind of text it is and pops a small context menu right there with the things you’re most likely to want:

• Highlight a word, run a search.

• Highlight a URL, open it.

• Highlight text, convert the case.

• Highlight an address, open it in Maps.

That’s it. It’s a context-sensitive way of acting on small pieces of text, and I use it so often that I forget it isn’t part of the operating system.

Readwise Reader

Part of being a knowledge worker now is keeping an archive of interesting things. I consume a lot of content across a lot of platforms, and my brain only rests once I know something is safely stored where I can get it back. Candidly, most of it I never return to. Knowing I could is the point.

Reader has a plugin for every browser. When I hit something worth keeping, one hotkey saves a copy plus a link to the original, lets me tag it, and lets me route it to Archive, Read Later, or Inbox depending on whether I want to see it again today. It captures emailed content if I forward it. It handles PDFs and full web pages natively. The search is fast and the hotkeys are good enough that my hands never leave the keyboard.

I’ve kept an archive like this for a long time, going back to the early Evernote days. What makes Reader better is the speed and the fact that it’s rarely missing anything. Every time I’ve sent the company a feature request, I’ve gotten a polite note back showing me where that feature already was.

The real unlock is the MCP connection. Reader is wired into my AI agents, so Claude can search my saved articles and my tags. Whenever I start a research project, one of the first places it looks is my own archive, filtered by the tags I’ve already applied. My reading history became a primary source instead of a graveyard.

Mylio Photos

In my line of work I deal with a punishing volume of photography: screenshots, samples of digital experiences, pictures of stores, pictures of shelves, pictures of things I’ve built. I have somewhere around 60,000 photos taken inside retail stores over the course of my career.

Mylio stores, syncs, and organizes all of it. It’s local and native, keeping multiple copies across devices I own, and it doesn’t require the cloud for anything if privacy matters to you. It puts optimized versions on my phone so I’m not hauling full-resolution files around, and if I need an original, the phone app pulls it down on request. I never think about any of this, which is the highest compliment I can pay a sync tool.

The organization and search tools are deep, and the AI categorization means I can usually find an image by typing the subject even when I never tagged it. It’s the best photo organizer I’ve owned.

Mylio also gets the customer relationship right. They have a team producing educational content and answering questions, a weekly webinar on features or use cases, and an unusually open development backlog. They’re in advanced preview of a major replatform right now with much deeper AI integration, and I’ve been playing with it and want to move my archive over.

The part that actually annoys me

Companies like Raycast and Mylio are so open about what they’re building, and so fast about shipping it, that they’ve ruined my patience for the legacy apps I’m required to use.

I have no idea if this is true, but I imagine there’s an enormous engineering organization at Microsoft working on Office. There are features I’ve used every single day for six years on mobile Outlook that still aren’t in desktop Outlook. I bounce between the web app, the iOS app, and the Mac app to get one job done, because certain things only exist in one of them. These are small annoyances I’ve been routing around for years, and they never get fixed, while new features I have no use for keep getting announced.

That would be irritating on its own. It’s worse when much smaller and far less resourced companies are so clearly doing it better. Raycast and Mylio are raising the bar, and some of the legacy apps are going to get left behind if they don’t notice.

Your turn

That’s the list. These tools make me happy, which is a strange thing to say about software, but there it is. I hope one or two are new to you.

What I actually want to know is what’s not on my list that you expected to be. Or what tool or habit gives you the most joy in your own day.

These 5 retailers captured 48% of all new spending in 2025.

1. Amazon $47.2 B
2. Walmart $22.8 B
3. Costco $16.9 B
4. TikTok Shop $6.1B
5. Home Depot $5.2B

Here is my best estimate of the top 5 retailers in the US in 2025, and what their GMV is. Note that this include Walmart only, not Sam’s Club $93B in US GMV, then Walmart Inc is the #1 retailers in the US in 2025.

Amazon $536 B
Walmart $495 B
Costco $ 200 B
Home Depot US $152 B
Kroger $148B

These 5 retailers alone represent 29% of all US retail sales (excluding Restaurants, Gas, and Automobiles).

His list was slightly historically inaccurate (due to how the source used to define retail), and hadn’t been updated in a while. So i made him (and you) a new one:

Or if you want to include GMV numbers in your version:

It now lives on a page that updates as companies report: retailgeek.com/comps

1. Retail is growing at two speeds again.

Across the 47 store retailers that reported a calendar Q1 2026 comp, the median was +2.9%. Across the 22 companies that disclose a digital growth rate, the median was +12.5%. That gap is the widest it has been since early 2021.

Digital had a real hangover after COVID. From mid-2021 through mid-2023 the median digital growth rate sat at or below zero in four of nine quarters and never topped 10%, while store comps stayed positive throughout. Digital only pulled away again in late 2024, and the median has climbed from about 6% then to 12.5% now.

2. Growth is broad again, at least through Q1.

In early 2024, barely more than half of tracked retailers were posting a positive comp (27 of 46 in Q1 2024). In Q1 2026 it was 40 of 47, or 85%, the broadest quarter since 2021. The Q2 read is still filling in (41 companies so far) and looks a notch softer: median +2.0%, with 27 of 41 growing. I’ll update the page as the September reporters land.

3. For a lot of retailers, digital is all of the growth.

Comps include digital. So a +1% comp next to +19% digital growth means the physical store business shrank. That is the pattern right now, and it is not just a couple of names:

• Walmart U.S.: comp +2.6%, e-commerce +24% • Kroger: identical sales +1.0%, digital +19% • Lowe’s: comp +0.2%, online +16% • Albertsons: identical sales -0.8%, digital +13% • BJ’s: comp +3.1%, digital +30% • Home Depot: U.S. comp +1.3%, online +11%

All most recent reported quarter, as reported by each company. Best Buy is the counterexample worth noticing: comp +4.5%, online +5.1%.

A word of caution before anyone puts these in a deck: every digital number is on the company’s own definition, and the definitions are all over the place. Walmart’s includes advertising and marketplace revenue. Some count store pickup as digital and some don’t. The page shows each company’s definition next to the number, because the definition is half the story.

Also: 14 of the 52 companies never disclose a digital growth rate at all, and a handful that used to have stopped. Digital disclosure in retail is still a courtesy, not a standard.

What to do with this Monday morning: if you benchmark your own comp against “the industry,” stop using a single average. Pick the six to eight retailers that actually compete for your customer and compare against them, on their stated definitions.

Which comp-versus-digital gap surprises you most? And who’s missing from the tracker?