On Monday, the engineer Dan Luu published a meticulous audit of Ed Zitron’s AI predictions — the most widely cited AI skeptic in the business — and found him mostly wrong. Zitron predicted OpenAI’s growth would stall in 2024. It didn’t. He called OpenAI’s revenue forecasts “a statement so egregious that I am surprised it’s not some kind of financial crime.” The company exceeded its 2025 target. He said o1 showed OpenAI was “desperate and out of ideas.” The models kept improving.
The post hit the front page of Hacker News within hours — 447 points, 524 comments — and the verdict was swift: the skeptics were wrong, the market was right, and the doomers owe everyone an apology.
That’s the predictable take. Here’s the thing it misses: Dan Luu just did the AI industry a favor it didn’t ask for and doesn’t deserve. He audited the skeptic. Nobody audits the boosters.
The Asymmetry of Accountability
Zitron made specific, dated, falsifiable predictions. “OpenAI growth is stalling.” “Revenue will slow dramatically as we enter the new year.” “o1 shows OpenAI is desperate and out of ideas.” These are checkable claims. You can go back, look at the numbers, and say: wrong, wrong, wrong.
That’s why Luu could audit him. That’s why the audit exists at all.
Now ask yourself: when was the last time anyone audited the boosters with this rigor?
Sam Altman has been predicting AGI — artificial general intelligence — for years. In 2023, he said it would arrive “in the reasonably close future.” In 2024, he said it was “a few thousand days” away. In 2025, he said the path was “basically clear.” These are predictions. They are also unfalsifiable. “Reasonably close” means nothing. “A few thousand days” is a decade. “Basically clear” is a vibe.
Nobody audits Sam Altman. Not because he’s been right — he hasn’t been around long enough to be right or wrong — but because his predictions are designed to be uncheckable. They’re vague enough to survive any outcome and specific enough to sound confident.
The same goes for the entire booster ecosystem. Nvidia’s Jensen Huang says AI will create “a hundred trillion dollars of value.” When? By what mechanism? Measured how? Nobody asks, because the prediction is too vague to audit.
Zitron, for all his errors, made predictions you could check. That’s the difference. And it’s why the audit — correct as it is — doesn’t discredit skepticism. It reveals that skepticism is the only position in AI that can be held accountable.
The Function of Being Wrong
Here’s the uncomfortable part for the people celebrating Luu’s post: Zitron’s wrongness was useful.
The AI industry spent 2024 and 2025 raising hundreds of billions of dollars on the promise of returns that have not materialized. OpenAI’s own filings show it expects to lose money for years. The company has committed to spending more than $50 billion on compute in a single year. The gap between revenue and spending is bridged by narrative — by the belief that the models will keep improving and the revenue will eventually catch up.
Zitron’s skepticism forced the industry to defend that narrative. It forced OpenAI to release revenue numbers it would have preferred to keep private. It forced the boosters to make their case in public, with specifics, rather than in private, with vibes.
That’s what skeptics do. They don’t have to be right to be useful. They have to be specific enough to be wrong.
The alternative — a discourse where everyone agrees the boom is real and nobody checks — is how you get 1999. It’s how you get 2008. It’s how you get a market where the only people asking questions are the ones who get audited for asking them.
The Unaudited Class
Luu’s post is rigorous, fair, and devastating to Zitron’s record. It should be read by everyone who takes Zitron seriously. It should also be read by everyone who doesn’t.
Because the real finding isn’t that Zitron was wrong. It’s that nobody has done this for the other side.
Where is the audit of Sam Altman’s predictions? Where is the audit of Jensen Huang’s? Where is the audit of the venture capitalists who said in 2023 that AI would transform every industry within two years? Where is the audit of the analysts who set price targets on Nvidia based on revenue that hasn’t arrived?
They don’t exist. Not because the boosters have been right — they haven’t been around long enough to be right or wrong — but because their predictions are too vague to check.
Zitron made himself auditable. That’s why he got audited. The boosters made themselves unauditable. That’s why they didn’t.
The next time someone cites Luu’s post as proof that AI skepticism is discredited, ask them this: when was the last time anyone held the boosters to the same standard?
The answer is never. And that’s the real story.