Anthropic guest post: Prof. Schwartz lets Claude find 'Claude-shaped' science problems
JMateosGarcia · x · 2026-10-02
Anthropic published a guest post by Harvard's Matthew Schwartz describing a new approach to AI-accelerated science. Instead of forcing Claude into traditional workflows, he let it find "Claude-shaped" problems suited to current LLM capabilities.
Key points:
- This led to BootLoops, a toolkit for exact calculations in quantitative science.
- Claude surfaced connections to ecology, population genetics and a dozen other fields—technically correct but often scientifically unremarkable—so Schwartz worked with domain experts to steer it toward questions those fields care about.
- His core observation: headlines about AI solving major problems clash with researchers' daily experience, since most breakthroughs sit in fully formalizable math, while most science doesn't.
- Designing problems around the model's actual capabilities is one way to bridge that gap.
Related event: Harvard Physicist Produces 36 Manuscripts with Claude in Three Months(13 posts)→
More from AGI Musings
- Critic pushes back: Anthropic leaking ideology into model weights is a red flag — gerardsans · 2026-11-16
- Alpha School accused of "counseling out" low-gain students as founders push back on selection bias claims — benjaminjriley · 2026-11-16
- Netflix lands Joseph Gordon-Levitt's AI thriller with Rachel McAdams, sparking debate on AI apocalypse blockbusters — ShakeelHashim · 2026-11-16
- Caine Prize confirms 2026 winner amid AI-use allegations — ShakeelHashim · 2026-11-16
- No legal liability framework exists for model misalignment harms, and that's a disaster — LuizaJarovsky · 2026-11-16
- A diagram for judging quality in cultural artefacts: coded meaning and exemplary moves — round · 2026-11-16