AI Solutions in Search of Problems

Normie consumer use case: Is consumer AI a solution in search of a problem?

Last year, Anthropic CEO Dario Amodei said that artificial intelligence could wipe out half of entry-level jobs in the U.S., spiking unemployment to close to 20 percent sometime in the next one to five years.

“Most of them are unaware that this is about to happen,” Amodei told Axios. “It sounds crazy, and people just don’t believe it.” 

Perhaps Amodei is talking about company-wide transformations as certain administrative, managerial, and tech jobs get automated away. The meme has long been “learn to code”—a snarky suggestion lobbed in the way of journalists who get laid off and bemoan the fact that their industry is less valued than it used to be, with less clear funding models than ever before. But what happens when learning to code is no longer good counsel? When even coders have been replaced by AI models? The conventional wisdom of the last five or 10 years no longer applies.

Still, compare these projections with the statistics we have on how many households nationally pay for AI subscriptions: Roughly 2.2 percent, as of April.

This is my thesis: no one uses AI. I repeat, absolutely no one. We live in a bubble.

Even among my friends who pay for it, when I ask them to open ChatGPT and show me their queries, it’s the same handful of basic things.

Most don’t even know they can upload a photo and ask… https://t.co/03GbUSJGKW

— Nicolas Bustamante (@nicbstme) October 3, 2026

The majority of consumers appear to currently use AI models as search engines: chatbots that can customize and tailor the search a little more finely, but not something meaningfully different. It’s a substitute for Googling.

Some smaller band of consumers—like Jesse Genet—use AI agents as personal assistants; there’s up-front work in training them to do such things, and there’s some amount of autonomy and trust extended, when you let them use your credit card to directly purchase things they believe you need.

But it’s not clear the second category is going to meaningfully overtake the first, even though it is perhaps the more sophisticated use case (and the one developers have in mind).

AI may be much more useful as an enterprise product than as a consumer product.

There’s a few important things going on here i think.

First, there’s a lot of people who aren’t ready for full leaps from doing lots of things themselves to having agents (or anyone for that matter) take a lot control from them. This is just bc people don’t like tons of change… https://t.co/MUE39SmsCS

— Charlie Warzel (@cwarzel) October 7, 2026

This is probably the correct take, at least as of right now:

The problem with a lot of AI assistant use cases is that specifying the work to be done in sufficient detail actually IS the work.

— Alan Cole (@AlanMCole) October 6, 2026

Of course, there are enterprise use cases that we haven’t even fully realized yet, that do hold great potential for massive disruption: coding agents that can build and maintain software with a fraction of the engineers it used to take; models that review contracts or process insurance claims faster than junior associates; customer-service systems that resolve problems rather than just routing them; and back-office tools that quietly absorb the scheduling, invoicing, and data entry that once justified whole departments. None of this requires people to trust an agent with a credit card. It only requires employers to decide the software is cheaper than the staff.

Then there are more innovative use cases: drug discovery, where models can screen millions of candidate molecules in the time it once took a lab to test a handful; protein-structure prediction, that has already handed researchers structural maps of proteins that once took years to work out one at a time; imaging, where models are being trained to flag tumors and early signs of disease missed by humans; and clinical trials, where AI could speed up the work of matching patients to studies and spotting safety signals in the data.

We shouldn’t let the discourse center around the comparative uselessness of the consumer products, when that’s not really where the value lies. But this probably also presents a public relations problem for these companies: It would be better for them if consumer-facing applications were wonderful, with clearly recouped value, as many people’s work lives will probably be disrupted by enterprise adoption of AI.

Scenes from New York: Ketamine, “which has surged in popularity in recent years, appeared time and again on the night in question, at a Chi Phi fraternity party in October 2024, according to the documents and a lawsuit filed last month by the former Cornell student who identified herself as Jane Doe,” notes The New York Times. The night of the alleged assault was Doe’s first time doing ketamine, per the lawsuit, and all seven of the named men apparently did ketamine as well. “The investigative files also show that on the night of the alleged assault, Jane Doe ran into a fraternity member she knew who was the D.J. of the party,” reports the Times. “Later on, she said another student told her that the D.J. had overdosed on ketamine and other substances the night before, the records show.”

Ketamine’s surge in popularity—which I talked about a bit on this week’s Reason Roundtable—is troubling but well documented. “From 2017 to 2024, the percentage of New York City nightclub attendees who reported having used ketamine has nearly doubled,” ketamine researcher Joseph Palamar told the Times. “Even on relatively small doses, you’re very detached from things.”

QUICK HITS

• “Billionaire Ray Dalio warned that Treasuries are vulnerable to a pullback of demand from China and Japan, two of the US’s largest foreign creditors, highlighting a potential pain point for a market that has endured wild swings this year,” reports Bloomberg.
• There’s a measles outbreak in Pennsylvania, and one pediatric infectious diseases specialist tells Leana Wen that this outbreak looks different than its predecessors: “[Patrick] Gavigan told me that his biggest surprise was how many severe cases he was seeing in previously healthy individuals, and how many involved non-respiratory symptoms,” writes Wen for The Washington Post. “Some patients experienced vomiting and diarrhea and were found to have elevated liver enzymes and ultimately diagnosed with hepatitis. Some developed encephalitis, an inflammation of the brain that can result in seizures, paralysis and permanent brain damage. Some have been discharged to go home; others have faced a much slower recovery and may never regain the abilities they had before becoming ill.”
• “Oura’s glossy smart ring is everywhere these days: in high-profile TV ad campaigns and on the fingers of plenty of people. Yet when Oura tried to go public late last month, it couldn’t find enough buyers at its proposed price and shelved the listing,” reports The Wall Street Journal. “The company blamed jittery markets, despite major stock indexes being near record highs, and said it could afford to pick its moment. Surging Treasury yields were certainly part of the problem. The bigger challenge, though, was a mismatch between how Oura sees itself and how prospective investors saw it.”
• Insane:

The cost of heating oil in Maine is up 80% from last year. https://t.co/0iD19D4PbG pic.twitter.com/x9Lqt1vwmx

— Joe Weisenthal (@TheStalwart) October 6, 2026