Dan Luu reviewed roughly 25 public predictions by the AI critic Ed Zitron and judged nearly all wrong. The predictions run from February 2024 to November 2025. Most claim that AI models had peaked, run out of training data, or hit a wall. Luu also disputes Zitron's 2024 claim that Meta, Google and Microsoft are dying. He sets out revenue and profit growth for all three through the first half of 2026. Other calls that failed include OpenAI revenue stalling, CoreWeave collapsing and Cursor dying. Luu says the underlying reasoning is weaker than the record, citing third-party traffic estimates and cherry-picked figures. He points to a spreadsheet error found by Timothy B. Lee in a projection of Anthropic revenue. Luu states he holds no financial stake in AI firms beyond index funds.
What changed
Zitron's forecasts were widely cited as numbers-backed evidence that AI progress had stalled.
What it unlocks
Checking a frequently cited skeptic's track record before repeating his claims.
- Meta 2025 revenue $201B, profit $83B
- Alphabet 2025 revenue $403B
- Gemini reached 750M users
- Cursor exit valued at $60B
Sources