Since the code went public, no version of Fluxer's rules for outside contributors has forbidden using an LLM privately to research, plan or check your own work. The rules on what you can submit changed a few times in 2026 and got stricter, and our AI stance page lists every version in the public history, with a link to each one.
Was using an LLM to learn always allowed?
Nothing in Fluxer's rules has forbidden it since the code went public on January 1, 2026, and at that point there was no written rule about LLMs of any kind. The contributing guide from March 4, 2026 said LLMs "can be helpful for rubber-ducking", which means explaining a problem out loud to someone so you can understand it better.
The LLM usage policy that contributors follow today, from August 12, 2026, says "An LLM may assist a contributor privately with learning, investigation, planning and review."
Could contributors submit code an LLM helped write?
From February 17 to August 12, 2026, the contributing guide let contributors submit code that an LLM had helped them write. The first rule, from February 17, 2026, said "You're welcome to use LLMs as a tool for automating mechanical work." It didn't ask for disclosure, and from March 4, 2026 the guide asked contributors to disclose any LLM use.
When did the rules get stricter?
On March 4, 2026, the contributing guide stopped people from using AI to write bug reports, pull request descriptions or comments, apart from a translation. From June 27, 2026, an AGENTS.md file told AI agents not to open pull requests or comment for anyone.
Since August 12, 2026, the LLM usage policy has said an LLM "MUST NOT author any part of a submission", even if you disclose it. Here's what it lets you use an LLM for.