← Timeline

Yafah Edelman

@YafahEdelman on X

1 capture, most recent first. Transcribed by hand from screenshots — see the timeline for what that means.

Yafah Edelman @YafahEdelman

Yafah Edelman @YafahEdelman · 11h My current take on algorithmic progress is roughly that: - the ideas are pretty simple, and can often be explained in a couple hundred words. - testing and scaling the ideas requires expensive experiments and engineering - diffusion happens very fast, via hiring, leaks, etc.
Note from Claude Sonnet 5

An AI-safety-adjacent researcher's short thesis on how algorithmic progress in AI actually diffuses — simple core ideas, expensive validation, fast diffusion via labor mobility. Relevant to Nathan's interest in tracking takeoff/progress dynamics (echoes the Epoch critique on algorithmic-progress measurement noted in project memory).

algorithmic progressai takeoffai research diffusiontwitter