The unit economics crisis in AI is real. AI companies optimise for two things: distribution, and unit economics. Distribution is working but the economics aren't. OpenAI's 2025 accounts, obtained by Ed Zitron and verified by the Financial Times, show about $34B in costs against $13B in revenue, and a loss bigger than the revenue. You can't pull a large upfront check out of a company that ended the year like that. Pushing harder doesn't put the money there. Which is why the form of the ask matters more than the size. A company that can't write a big check can still pay for what it uses, every time it uses it - that‘s monetization at grounding…

Two weeks of watching this argument: everyone agrees publishers should be paid. Nobody agrees on the unit. Per deal, per crawl, per citation, per grounding. The unit is the whole fight.

We built one and patented it. It's being tested with AI companies and content providers now. Biased, obviously.

Cloudflare built pay-per-crawl. Now it calls crawling "a crude measure of value" and is moving to pay-per-use: answer engines pay when your content actually gets used. Google's doing the same. Good. It's the direction I've argued for over ten years. One catch though. A cover charge just became a performance fee, and performance pay is how advertising works... someone else keeps the scoreboard. Publishers should set the price, not just get one.

Here's a thing that bugs me. Every market runs on the deals that closed. Buyers get what they want, sellers get paid, and that's what gets recorded. An agent looks at your price, decides it's too much, walks away. Just like a human. But a human leaves a trace, like an abandoned cart, a session that stops at the price page. An agent leaves nothing; no transaction, no record, nothing in the logs.... you never find out that someone wanted it at all. That's demand at a price you didn't get, and it's how you'd learn what your work is actually worth. I think right now it's the one number nobody has and everybody needs for a market to form. Mastercard's agentic tokens capture intent, and that part is starting now. But if you captured intent plus the price it died at, you could stop guessing at prices and start testing them. You move the number and see who still walks... Log intent when the purchase happens and log it when it doesn't.

Dario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement. I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible. But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want. Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders’ “shut it all down.” China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well. So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you’ll buy goodwill for the next conversation. If you don’t, we’ll know this was just another bid for regulatory capture — or an election-season psyop.

Here is a neat use case for Grok Bot: Friends said The Odyssey in IMAX was sold out for weeks. I told Grok Bot to check every 30 minutes. Someone returned tickets, Grok Bot pinged me and I got my seats. No magic prompt, just simple language. Well done @bot

Apparently, I’m buying CDs again. Because... 19-year-old car. No Bluetooth. No CarPlay. Anyone still remember the feeling of enjoying one album at a time? 💿

Training is a lost cause for publishers. That's gone. The real opportunity and leverage: LLMs need fresh, accurate content for grounding, Supertab CEO @cosmoene told me. "An LLM can tell you all you need to know about molecular biology, but it cannot tell you if the restaurant next door is open."

You have noticed it. ChatGPT feels dumber than it used to. Your prompts that worked six months ago produce worse results now. The writing sounds flatter. The ideas sound safer. The internet itself feels like it is shrinking. Every article reads the same. Every email sounds the same. Every answer sounds like it was written by the same voice. You thought it was you. It is not you. Researchers at Oxford and Cambridge published a paper in Nature proving what is happening. They call it Model Collapse. Here is the mechanism in one sentence. AI trained on AI-generated data gets dumber every generation until it forgets what real human data looked like. The internet is filling with AI-generated content. Blog posts. Articles. Reviews. Comments. Social media. AI companies scrape the internet to train the next generation of models. Which means the next generation of AI is being trained on the output of the current generation. Each cycle loses information. Not randomly. It loses the rarest, most unusual, most creative parts first. The researchers call these the "tails of the distribution." The weird ideas. The unexpected perspectives. The things that made the internet feel human. Those disappear first. What remains is the average. The safe. The expected. The bland. Then the next generation trains on that. And loses more. And the next generation trains on that. And loses more. The researchers proved this is not a slow decline. Major degradation happens within just a few iterations. Even when some of the original human data is preserved. They tested it on large language models. On image generators. On statistical models. The pattern was the same every time. The output converges toward a narrow, flattened version of reality that looks nothing like the original data. The lead researcher put it plainly. "Large language models are like fire. A useful tool. But one that pollutes the environment." The pollution is invisible. You cannot see which sentence on the internet was written by a human and which was written by AI. Neither can the AI that is about to train on it. And once the tails are gone, they do not come back. The damage is irreversible. This is not a prediction anymore. It is a diagnosis. The internet you grew up on was built by humans writing things no algorithm would have written. Strange, personal, imperfect, alive. That internet is being diluted. One generation of AI at a time. And the models trained on what remains are learning a smaller and smaller version of the world. Model Collapse is not a technical problem. It is a cultural one. The thing that made the internet worth reading is the thing that disappears first.

The one thing a bot can’t consume, even if it wanted to pay for it: a donut wall. 🍩🍩🍩Great catching up with the Fastly team at Xcelerate in London. Glad to see the conversation moving past ‘AI is coming’ to the real questions - new business models (with Supertab as part of it), bot traffic, securing the stack. @fastly #xcelerateldn2026 @deflatermouse