Jason Wei on 'verifier's law': the asymmetry of verification is now a core AI idea
_ScottCondron · x · 2026-10-27
Scott Condron resurfaces a blog post by OpenAI researcher Jason Wei on asymmetry of verification. With RL finally working in a general sense, the idea that some tasks are far easier to verify than to solve is becoming one of the most important ideas in AI.
Wei categorizes tasks into three buckets:
- Much easier to verify than solve: sudoku/crosswords; coding an Instagram-scale website (engineers-years to build, a layperson can verify quickly); BrowseComp problems (browsing 100 sites to solve, fast to verify).
- Near-symmetry: adding two 900-digit numbers, or some data processing scripts—verifying takes as long as writing.
- Harder to verify than solve: fact-checking an essay versus writing it (Brandolini's law), and many scientific hypotheses.
More from AGI Musings
- AI pundit predicts the 2020s will be the last decade of widespread poverty — Dr_Singularity · 2026-10-28
- Essay: your bodymind runs on --dangerously-skip-permissions, and you never opted in — mimi10v3 · 2026-10-28
- AI labor paper: Amdahl's Law explains who gets hit and who's shielded — not just the young — soumitrashukla9 · 2026-10-28
- LAIKA director vows never to use generative AI in films: "AI is theft" — moultano · 2026-10-28
- Ethan Mollick: the "AI is all fake" conspiracy should have died, but won't — emollick · 2026-10-28
- Google researcher: refusing AI as an artisan is honorable but impractical for the economy — moultano · 2026-10-28