Ed Zitron's AI predictions do not hold up under fact-check
An unnamed author writing on danluu.com set out to check how well the forecasts of Ed Zitron, described as the most widely cited AI skeptic the author has seen, have actually panned out. The author states having no financial stake in AI companies beyond standard index-fund exposure and no history of a fixed pro or anti AI position, then walks through Zitron's public statements and finds the predictions and the reasoning behind them wrong.
The centerpiece is a November 2024 talk in which Zitron called Meta "a dying product, and it's kind of a dying company" and said Google and Microsoft, alongside Meta, no longer know how to grow, which is why the tech industry is pushing AI into everything out of desperation. The author checked this against the three companies' actual revenue and GAAP operating income and found all three growing, not dying. Zitron's evidence for a Facebook decline came from third-party Similarweb estimates rather than Meta's own reported user numbers, which the author says show no sustained drop. For Google, Zitron blamed executive Prabhakar Raghavan, calling him "a computer scientist class traitor that sided with the management consultancy sect," for damaging Google search and undermining the work of former Google engineer Ben Gomes. The author notes Google search engineers who commented on the claim did not agree Raghavan was the main cause of any search problems, and that even a damaged search product would not stop Google from growing given other revenue lines such as YouTube and Google Cloud.
A separate example concerns Anthropic. Commenter Timothy B. Lee reviewed a spreadsheet Zitron used to project Anthropic's 2025 revenue and reportedly found it miscounted dates: it skips February 1 to 10, counts March 1 to 10 twice, treats August 21 to October 21 as a single month instead of two, omits October 21 to November 1; another commenter separately noted the spreadsheet also contains a nonexistent February 30. Zitron had used the spreadsheet to arrive at a $3.6 billion revenue projection and suggested the discrepancy pointed to "funny business" in Anthropic's reporting; Lee's correction of the date errors reportedly makes the numbers consistent instead. The author also quotes commenter Juho Snellman, who argues Zitron's articles link to sources that, followed to the primary material, say something different from what Zitron implies, and calls Zitron's economic analysis "absolute trash" that looks like "intentional deception" given the consistent direction of the errors.
The piece then lists a run of Zitron predictions from February through July 2024: that generative AI was reaching the upper limits of what it can do, that AI had hit "peak AI," that AI companies were running out of training data and could not improve further, and that OpenAI's growth was stalling in a way that would lead to some kind of collapse. The author marks each as wrong. The excerpt provided ends mid-list, before the article's final tally of how many Zitron predictions were checked in total.
Key facts
- A danluu.com post checks specific 2024 predictions and claims by AI skeptic Ed Zitron against actual data and finds both the predictions and the underlying reasoning wrong.
- In a November 2024 talk, Zitron called Meta a "dying product" and "dying company" and said Google and Microsoft also do not know how to grow, but the author's review of the three companies' revenue and profit shows growth, not decline.
- Zitron's Facebook decline claim relied on third-party Similarweb estimates rather than Meta's own reported user numbers, which the author says show no sustained drop.
- Commenter Timothy B. Lee found the spreadsheet Zitron used to project Anthropic's 2025 revenue at $3.6 billion contained date-counting errors; another commenter separately noted the spreadsheet also contains a nonexistent February 30; correcting the errors makes the numbers consistent.
- The author lists Zitron predictions from February to July 2024, including "peak AI" and a stalling-OpenAI-leads-to-collapse claim, and marks each one wrong; the retrieved excerpt ends before the piece's final scorecard.
Why it matters
Zitron is cited widely in AI discourse as someone who has "looked at the numbers," and the author argues that citation functions as borrowed authority rather than a reflection of sound analysis. The post matters because it pushes back on a specific, frequently repeated narrative, that major tech companies are dying and turning to AI out of desperation, by checking it against public revenue and profit figures rather than rhetoric. The story drew 518 points and 609 comments on Hacker News within about 10 hours, indicating the underlying debate over AI-skeptic commentary is live and contested.
Who it affects
Readers who follow AI industry commentary and cite Zitron as a data-backed voice, other commentators and journalists who reference his claims about Meta, Google, Microsoft or Anthropic, and, indirectly, Zitron himself, whose spreadsheet and rhetoric are the direct subject of the critique. It also affects anyone forming views on whether large tech companies are struggling and pushing AI as a growth substitute.
How to use it
The post is a model for checking a public commentator's claims against primary data rather than accepting rhetorical framing: compare a claimed decline against a company's own reported revenue and user figures rather than third-party estimates, and check a cited spreadsheet's underlying math (as Timothy B. Lee did with the date-counting errors) before repeating a number pulled from it.
How solid is it
The author cites specific, checkable figures: company revenue and GAAP operating income for Meta, Alphabet and Microsoft, a named commenter's (Timothy B. Lee) line-by-line audit of a Zitron spreadsheet, and another named commenter (Juho Snellman) with a documented history of challenging Zitron's claims in Hacker News comments. The author discloses no financial stake in AI companies beyond standard index-fund holdings. The retrieved text does not include the piece's final tally of predictions, its full source citations, or any response from Zitron himself.
Risks and caveats
The retrieved excerpt cuts off mid-list, before the article's stated conclusion or a final scorecard of right versus wrong predictions, so the full extent of the case cannot be verified from this text alone. The critique rests on one author's reading of Zitron's talks and articles plus quoted third-party commenters; the piece does not quote or characterize any response from Zitron to these specific criticisms. The author's own name is not given in the retrieved text.
“it's a dying product, and it's kind of a dying company”
— Ed Zitron, in the November 2024 talk examined by the post