🔬 An Oscar, Two Asteroids, and the Algorithm in Your sklearn: John Platt on AI for Science

Google's ERA shows LLM-guided evolutionary search became scientifically useful only at Gemini 2.5, unlocking real climate-science breakthroughs.

• Google's ERA system: Empirical Research Assistance uses Gemini in a Monte Carlo Tree Search loop with Upper Confidence Bound selection, mutating ~10 notebooks per iteration to maximize a 'scoreable task' function.
• Step-change at Gemini 2.5: Platt says ERA-style approaches simply didn't work under Gemini 2.0 but became powerful with 2.5, calling the tool 'like having a hyper-eager grad student who doesn't sleep.'

Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.