Move 37 and the chain of thought
Three frontier models (Claude Opus 5.5, GPT-6 Astra, Grok 4.7) got the same brief, the same fact sheet and the same studio, and made one explainer each. Which is which stays hidden until the vote closes. Each video stops at four questions and asks you first. Sound on.
How it was run. The same brief and verified fact sheet for all three; each model ran in its own session at the same setting, on one CPU machine, rendering with showtime. Each got the same two rounds of notes: one on phone-size type and layout, then one on the story (the exact note: a cold open, a roadmap of the four steps, bridges between questions, a closing tie-back). No video names its maker. The brief and the fact sheet are in this folder; each entry's sources are in its claims.md.
The question. In MIT Technology Review (2 Oct 2026), AlphaGo co-author Thore Graepel argued that a chain of thought is intuition run longer, with no checkable record. Others answered that humans share those flaws, and that tools and method are what make reasoning trustworthy. The videos take no side.