How accurate have Ed Zitron's AI skeptic predictions been?
TL;DR
Dan Luu checks about 30 dated predictions the AI skeptic Ed Zitron made in 2024 and 2025. They include that AI had peaked, that OpenAI's revenue forecasts were near-criminal, and that Cursor would die. Luu scores nearly all of them wrong. He says Zitron's numbers decorate anger more than they support an argument.
The article
- Luu opens with Zitron's November 2024 claim that Meta, Google and Microsoft were dying and pushing AI out of desperation. All three grew revenue and profit by double digits from 2023 through the first half of 2026. The supporting evidence was an outlier Similarweb estimate of Facebook users and a hero-villain story about Google's Prabhakar Raghavan.
- People cite Zitron as "the guy who looked at the numbers". But the numbers often don't connect to his claims. Timothy B. Lee found that the spreadsheet behind Zitron's Anthropic revenue estimate counted March 1-10 twice, among other date errors. Another commenter spotted a February 30 in it. Once the errors were fixed, the suspicious gap vanished.
- The dated list runs from February 2024 to November 2025. Zitron said generative AI had peaked, labs were out of data, and Cursor would die. He called OpenAI's $11.6B forecast for 2025 "some kind of financial crime". He said Gemini reaching 500M users was so unrealistic that Sundar Pichai should be fired. He put the bubble's pop "no later than Q2 2026". OpenAI beat the forecast, Gemini hit 750M, and Cursor sold for $60B. Only the Orion-to-GPT-4.5 call rates "possibly right".
- Luu stresses confidence. Zitron is wrong at the highest rhetorical certainty possible. Under Brier scoring, his record would score badly. Luu left out near-tautologies such as "OpenAI will collapse or raise more money". A redditor's earlier tally had counted that one against Zitron.
- He compares Zitron to Kurzweil, who also uses numbers for an aura of credibility. He also compares him to Paul Ehrlich, who never conceded his famine predictions. Following Michał Zalewski, he argues that crisp positions that leave no room for doubt win audiences. He lists the defenses fans use: "you didn't address X", "he never said that", "he's just early".
Why it matters
Worth reading if you cite Zitron or are weighing AI-bubble arguments. The original links every prediction to its source and quotes Juho Snellman and others on Zitron's use of numbers. Its appendices explain the method: ChatGPT drafted the prediction list, Luu filtered it, and he says which claims he left out and why. It does not take on the data-center financing case that most of Zitron's readers follow him for.
The discussion
Across over a thousand comments, most commenters accepted that the dated predictions failed and fought hotly over whether that matters. One side called Zitron a grifter captured by his audience. A loud minority said Luu graded slogans literally and wrote "Wrong" without reasons. They read Zitron for the financial argument about circular deals and data-center debt, which Luu skipped.
Wrong, or just early?
The longest argument, which one commenter framed as strict graders against people who care about direction, was over whether a missed date still matters. Defenders compared Zitron to those who called the dot-com and subprime busts years early, and said causal analysis matters more than timing. Critics answered that he chose the deadlines himself, and that "AI has already peaked" was false when he said it. A later pop, they said, would not fix the record.
"Early" works for an open-ended bubble thesis. It does not rescue dated claims that already failed.
Some people in this thread are like "well the value is in the prediction", but from my perspective I'm thinking "the value is in the causal analysis".
The financial thesis the post skipped
Many said Luu dodged Zitron's main case: AI revenue is tiny next to capex, usage is subsidized, and money cycles between hyperscalers, labs and startups. Commenters cited Harvey's token bills, power users running 15 Codex plans, Google booking gains on its Anthropic stake, and off-balance-sheet data-center debt. Others pushed back: paying customers are real, and investment gains sit apart from the revenue lines Luu quoted. Nathan Tankus, they noted, argues AI bond issuance is too small to cause a subprime-style crisis.
the whole thesis is XYZ likely fail because REVENUE RECORDS is not enough to dig out of hole relative to MAGNITUDE MORE SPEND.
Yes the investments do increase GAAP, but these are seperate line items from revenue which is what is listed in the article.
Critiques of Luu's method
Several readers found the grading uncharitable, with bare "Wrong" verdicts on fuzzy claims like "peak AI". They said revenue tables can't refute "dying" when Zitron meant enshittification, and Luu himself agrees Google degraded search for revenue. They also faulted Luu for taking Gemini's 750M users at face value after debunking futurists' marketing numbers. Others replied that Zitron made the claims in those words, and rigor is the norm when grading predictions.
it doesn't even attempt to interpret the claims in any way except the least favorable.
Luu also fixates on Zitron mentioning Prabhakar Raghavan, but then proceeds to agree with Zitron's core point that Google has intentionally degraded it's search product to maximize revenue.
Have models improved since 2024?
Engineers who use the tools daily called the "capabilities have peaked" claims absurd and described going from a toy to their most improved tool. A smaller group credited the gains to harnesses, agents and coding-focused training, not the models themselves. One commenter claimed GPT-3.5 in a modern harness would match Opus 5, and most replies treated that as indefensible.
Over that time it has gone from being an interesting toy that could kind of help but would often hinder, to being an absolutely explosively powerful tool.
The distinction I would like to make is that it's not the models that are improving, but rather the infrastructure around them.
Incentives, audience capture and the CEOs
Many saw a pundit captured by a paying audience that wants AI to fail, and some had firsthand stories. minimaxir stopped helping him over misused advice, gregdoesit was mocked and blocked after sharing survey data, and a former PR client found him difficult. simonw said Zitron sometimes breaks real news but buries it in commentary too biased to judge. Asked why Altman and Amodei get a pass, most answered that the CEOs also build products, while predictions are Zitron's whole output.
I stopped assisting him because he repeatedly misused said advice to the most cynical interpretation
What else is Zitron doing that he could be evaluated on, besides making predictions that turn out to be wrong?
Consensus
Zitron's dated calls on capabilities, revenue forecasts, Gemini users and Cursor failed. His tone and audience make him a poor source. Even many who share his bubble worries wanted a more careful critic.
Dissent
A sizable minority held that the core claim is still open: AI revenue can't cover the capex and circular financing. They found Luu's piece as one-sided as he says Zitron is, with "Wrong" in place of reasons.