{"slug": "harvard-physicists-bootloops-toolkit-puts-ai-to-work-across-the-sciences", "title": "Harvard physicist’s BootLoops toolkit puts AI to work across the sciences", "summary": "Harvard theoretical physicist Matthew D. Schwartz released BootLoops 1.0 on October 1, 2026, an open-source, model-agnostic toolkit that pairs large language models such as Anthropic's Claude with scientific problems he calls \"Claude-shaped\" tasks. Over three months, Schwartz and 19 coauthors produced 36 manuscripts from a pool of 400 candidate problems, and the toolkit's flagship use case computed 30 mathematical physics integrals end-to-end, 15 reproducing known results and 15 new, including elliptic Feynman integrals. BootLoops ships under the MIT License and is framed as research instrumentation, not for clinical, actuarial, regulatory or safety-critical decisions.", "body_md": "Anthropic official brand assets (anthropic.com)\n\n# Harvard physicist’s BootLoops toolkit puts AI to work across the sciences\n\nMatthew Schwartz's open-source harness matches large language models to the scientific problems they handle best\n\nHarvard theoretical physicist Matthew D. Schwartz took a more practical route. On October 1, 2026, he released **BootLoops 1.0**, an open-source toolkit built on one premise: AI works best when you give it the right kind of problem.\n\nSchwartz calls these problems “Claude-shaped” tasks. Over three months, he and his collaborators used that framing to turn out 36 manuscripts with 19 coauthors, drawn from a pool of 400 candidate problems.\n\n## What BootLoops actually does\n\nBootLoops is a modular harness for scientific calculation. It sits between a researcher and a large language model, such as [Anthropic](https://cryptobriefing.com/markets/anthropic/)’s Claude.\n\nThe flagship use case is mathematical physics. The toolkit can compute 30 integrals end-to-end.\n\nHalf of those, 15 integrals, reproduce results that were already known. The other 15 are new, and they include elliptic Feynman integrals.\n\nFor non-physicists, Feynman integrals are the math behind how particles interact in quantum field theory. The elliptic kind are notoriously hard to evaluate.\n\n## Well beyond physics\n\nThe project does not stop at particle physics. BootLoops reaches into roughly 18 to 22 different fields, including ecology, population genetics and cosmology.\n\nVerification is central to the design. BootLoops produces its outputs as Python scripts, which return results to arbitrary precision.\n\n### AI, tech, and the markets they move—in one daily briefing.\n\nDaily. Free. Join 34,000+ readers across crypto, finance, and policy.\n\n## How it was built\n\nBootLoops grew out of an iterative process with Claude that began in December 2025. Schwartz started with Claude Opus 4.5.\n\nBy the summer of 2026, the workflow had moved to Claude Fable 5.\n\nThe launch followed a guest post by Schwartz on Anthropic’s blog titled “Claude-shaped science.” Anthropic published the piece.\n\nDespite its origins, BootLoops is model-agnostic. It is designed to work with other LLMs, including [Google](https://cryptobriefing.com/markets/alphabet/)’s Gemini and [OpenAI](https://cryptobriefing.com/markets/openai/)’s ChatGPT.\n\nThe software ships under the MIT License.\n\n## Who is behind it\n\nSchwartz is a Professor of Physics at Harvard. His work spans quantum field theory, particle physics and machine learning.\n\nHe also wrote a textbook on Quantum Field Theory.\n\nThe toolkit is explicitly framed as research instrumentation. It is not intended for clinical, actuarial, regulatory or safety-critical decisions.\n\n## What this means for research\n\nThe most useful idea in BootLoops may be the selection step, not the software. Starting from 400 candidate problems and producing 36 manuscripts implies a lot of filtering.\n\n**Disclosure:** This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our\n\n[Editorial Policy](https://cryptobriefing.com/editorial-policy/).", "url": "https://wpnews.pro/news/harvard-physicists-bootloops-toolkit-puts-ai-to-work-across-the-sciences", "canonical_source": "https://cryptobriefing.com/bootloops-toolkit-ai-scientific-research/", "published_at": "2026-10-01 19:02:18+00:00", "updated_at": "2026-10-01 19:19:09.039492+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "ai-tools", "ai-products"], "entities": ["Matthew D. Schwartz", "Harvard University", "BootLoops 1.0", "Anthropic", "Claude", "Claude Opus 4.5", "Claude Fable 5", "Google Gemini"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/harvard-physicists-bootloops-toolkit-puts-ai-to-work-across-the-sciences", "markdown": "https://wpnews.pro/news/harvard-physicists-bootloops-toolkit-puts-ai-to-work-across-the-sciences.md", "text": "https://wpnews.pro/news/harvard-physicists-bootloops-toolkit-puts-ai-to-work-across-the-sciences.txt", "jsonld": "https://wpnews.pro/news/harvard-physicists-bootloops-toolkit-puts-ai-to-work-across-the-sciences.jsonld"}}