Dispatch
Dan Luu scored roughly 25 of Ed Zitron's AI predictions and found essentially none of the major ones came true
Luu's September 1 post, which drew 852 points and over 1,000 comments on Hacker News, walks predictions from February 2024 through November 2025 after removing unfalsifiable and tautological ones. The misses include "AI has peaked" (Feb 2024), Meta "dying" (Nov 2024) against revenue going from $135B to $201B between 2023 and 2025, OpenAI's $11.6B 2025 revenue target being absurd, Gemini's 500M-user goal being unrealistic against an actual 750M, and the bubble popping by Q2 2026. Luu's sharper argument is that the reasoning chains are broken even where a call happened to land, which is the part worth borrowing when you evaluate any AI forecaster.
↳ Follow the thread