Anthropic researcher's three-bucket framework for doing science in the AI era
CatAstro_Piyush · x · 2026-09-23
- SGRodriques lays out a three-bucket framework for deciding what science to do now that AI is transforming research:
- Fast, easy-to-verify problems — a small subset of all problems, but the ones AI will impact most; their published Millennium Problems are an example.
- Problems that make other problems verifiable — e.g., speeding up small-molecule synthesis unlocks many more problems; similar: new assays, biomarkers, faster clinical trials.
- Gathering data for the truly hard problems — in the age of AI, data is king.
More from AGI Musings
- Calls for US-China AI governance mechanism date back a year, author notes — pstAsiatech · 2026-09-23
- Is AI-fatigue real? Redditors vent over endless hype vs. mediocre reality — Lvceateisdomine · 2026-09-23
- Why is every AI product still a generic chat box? — Complete_Snow_6437 · 2026-09-23
- "Stochastic Parrot" Feud Resurfaces as LLMs Outdo Humans at Frontier Math; Oversight Debate Rekindled — ctjlewis · 2026-09-23
- The best startup founders are cult leaders — ai · 2026-09-23
- Lists are the clearest AI writing tell: plausible at a glance, hollow on inspection — sethlazar · 2026-09-23