OCaml maintainers reject massive AI-generated pull request (devclass.com)

OCaml maintainers rejected a 13,000+ line pull request generated with Anthropic’s Claude Code that attempted to add DWARF debugging support to the OCaml native compiler (ocamlopt). The submitter, Joel Reymont, said he “shepherded” the AI rather than writing code by hand; the output included references crediting a Jane Street engineer and bore resemblance to OxCaml’s existing DWARF work. Maintainers cited copyright and provenance concerns, the absence of a prior design discussion, the massive review burden, and overlapping in-progress work as reasons to close the PR—noting that OCaml’s project organization and reviewer capacity aren’t set up to absorb huge, AI-produced changes. The episode matters for the AI/ML community because it highlights practical risks of automated code generation: attribution errors and possible copying from existing codebases, nondeterministic LLM outputs, and a scaling mismatch between PR volume and human review capacity. Technically, the patch showed an LLM can produce intricate, working features (DWARF is the standard used by gdb/lldb), but maintainers find reviewing AI-produced patches more taxing than human-written ones. The incident underlines the need for clearer open‑source policies on AI-assisted contributions, better provenance tools, and processes to prevent review bottlenecks as AI-driven PRs proliferate.