What a framework for mathematics done with AI has to do, what already exists, and a modest proposal for one part of it.
human-ai-mathematics.github.io
AI systems now produce mathematics in quantity, and the community has answered with many projects, each addressing one part of the problem. What is missing is a view of the whole: what a framework has to do, which parts are well served, and where nothing exists yet.
What a framework for mathematics with AI has to do, and about 150 projects mapped against it, layer by layer, from sources to communication (survey dated 21 October 2026). Three architectures compares ways of connecting what exists: an integrated platform, a federation, or a registry for research records.
A shared, certified record of a sustained human–AI search on one hard statement, in which every status is certified by an independent review or a named person. How it works: the life of a statement, the repository, and how to contribute.
- The KLS theorem and its methods — our testbed. KLS was proved three times in October 2026, while our search was running; the manuscript became an account of the three proofs, their comparison, and alternative mechanisms for the dimension-free Poincaré bound.
- conjecture-search-template — the reusable research harness behind it: a MyST manuscript, a ledger of certified claims, and researcher, reviewer and writer agents. It can start a search on any statement.
The site is meant to be temporary: this work makes most sense as part of a larger initiative, and the site will point there when one exists. Until then, start a discussion or open an issue, or write to nicolas.brosse [at] ensae.fr.
Nicolas Brosse · Jia Li · Yuwei Lyu
Supported by Project Numina.
Nicolas Brosse works at CREST, ENSAE Paris, Institut Polytechnique de Paris, with Arnak Dalalyan. His work is funded by the European Research Council (ERC) under the European Union's Horizon Europe research and innovation programme (grant agreement No. 101201229).