Hexagon Launches Repository for LLM-Assisted Mathematical Research

The mathematical and theoretical computer science communities have established a new archival platform, Hexagon, designed to host and cite research increasingly assisted or generated by large language models. Launched in early August 2026, the repository was developed by a coalition of academics and practitioners who identified a critical gap in scholarly infrastructure following the rapid proliferation of AI-generated mathematical discoveries. Traditional preprint servers like arXiv have maintained their existing submission standards and declined to create a dedicated channel for AI-assisted work. Meanwhile, fragmented hosting on GitHub repositories and academic social networks has proven inadequate for reliable long-term citation and preservation. To address this, the group founded the Hexagon Mathematics Foundation, a Delaware-registered nonprofit currently applying for 501(c)(3) tax-exempt status. Ben Antieau serves as the foundation's inaugural Executive Director, with a board comprising Mohammed Abouzaid, Francois Charles, Bryna Kra, David Savitt, and Lauren Williams, alongside an advisory panel featuring prominent mathematicians including Kevin Buzzard and Ravi Vakil. Hexagon operates under a distinctly permissive policy framework tailored to the AI research cycle. The platform explicitly accepts submissions entirely generated by language models, including works where contributors claim limited human oversight or understanding. It also accommodates incremental findings and micro-results, provided they hold academic interest. Unlike formalization-focused initiatives, Hexagon prioritizes accessibility and community discovery, linking directly to established formal proof ecosystems such as mathlib and TauCeti. The repository currently operates in beta and is privately funded by its initial contributors, though the foundation is exploring sustainable revenue models. Operational controls include tiered submission rate limits that scale based on moderation performance, alongside full API access for both read operations and authorized submissions. The platform's name derives from Jorge Luis Borges's conceptual Library of Babel, symbolizing an organized archive for the expanding volume of algorithmically derived knowledge. Antieau, who previously documented an LLM-collaborated discovery of derived invariance properties in characteristic zero fourfolds, emphasizes that Hexagon aims to capture mathematical insights at their earliest viable stage. The vision is for the repository to serve as a foundational layer where AI-emergent results gain community validation before advancing to traditional preprint and peer-review channels. As large technology firms continue to produce high volumes of computational and theoretical proofs, Hexagon positions itself as the community-maintained standard for citation, preservation, and open access in the emerging era of algorithmic mathematics.