GoKawiilPhysicist builds BootLoops toolkit after letting Claude pick its own math problems
Prof. Matthew Schwartz describes developing BootLoops, a toolkit for exact calculations in quantitative science, after changing his approach to working with Claude. Rather than directing Claude toward specific physics problems, he let the model identify calculations suited to its own capabilities, which surfaced connections across fields including ecology and population genetics. Schwartz then collaborated with domain experts to refine these findings into questions those fields actually find meaningful.
GoKawiil's interpretation of the reporting above, not reported fact.
The account suggests current AI models may contribute most to science not by replicating human research processes but by finding cross-disciplinary mathematical patterns that experts can then shape into useful questions. This implies a collaborative workflow where human domain expertise remains necessary to translate technically correct AI output into scientifically relevant results. Schwartz's experience could inform how other researchers structure their own use of LLMs for research rather than treating them as autonomous scientists.
- Schwartz built BootLoops, a toolkit for exact calculations, after adapting his workflow to Claude's strengths.
- Claude identified mathematical connections across fields like ecology and population genetics, though initially unremarkable to specialists.
- Domain expert collaboration was needed to turn Claude's technically correct findings into questions relevant to each field.
Source: anthropic.com, 2026-10-02
Published there as: “Claude-Shaped Science”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.