LLVM Developers Begin Debate Over AGENTS.md For Helping AI Agents

LLVM developers have begun a discussion on whether the project should host an AGENTS.md file (and CLAUDE.md) in their Git repository for helping to guide AI agents. While many open-source projects have adopted AGENTS.md files for helping to guide AI agents working on patches or other work around open-source contributions, not everyone is in agreement over their effectiveness or just how welcoming to be towards AI agents. With LLVM there are concerns mentioned too given the number of sub-projects involved in LLVM with different scopes. Longtime LLVM developer Nick Desaulniers started the discussion yesterday on LLVM Discourse whether to pursue the AGENTS.md/CLAUDE.md additions with the responses so far being mixed. While some are in favor of it, some find that these files can actually lead to more noise and not necessarily effective. Some also expressed that it goes against current LLVM principals of not shaping one developer's private workflows. Some also are already using their own custom AGENTS.md material in their own private repositories. Nick did open a draft pull request for adding a minimal AGENTS.md to LLVM. With that minimal version it tells AI agents to review the LLVM documentation concerning AI tool contribution policy and on their coding standards. We'll see what position LLVM ultimately decides to take on AGENTS.md or not.