AI Code Rating – a 3-character disclosure of how AI was used in a repo

AI Code Rating is a draft proposal for disclosing, in three characters, how AI was used to build a software project (aicoderating.com). Its author introduced it on Hacker News as a quick way to tell how much an open-source project relies on AI (Hacker News). A maintainer picks a rating and saves it, with a short plain-English note on how AI was used, in a file called ACR.md at the root of the repository.

Each character answers a separate question, and each uses a different kind of symbol so a rating cannot be misread. An uppercase letter from A to E rates the expertise of the person responsible for the code; on a team, that is the least experienced person who can approve a change alone. A digit from 0 to 4 gives the share of the code, by lines, that AI generated; single-line autocomplete does not count, but accepted multi-line suggestions do. A lowercase letter from a to e says how closely a person checked AI-written code before it was merged, and review by another AI does not count. The author's DomainMOD project is rated B1b: a maintainer who works professionally in that stack, AI-written code making up 1 to 25 percent, and every AI change read by a person before merging. The site rates itself B4b, meaning AI wrote nearly all of it under review.

The site argues that keeping these three questions apart is what sets it apart from disclosure schemes that use a single level, so that, in the site's example, a project rated A3a can be stronger than one rated C1e. It also says the rating does not judge whether using AI is good or bad. The limit is that ratings are self-reported by maintainers. The site offers a validator and a GitHub Action that check the ACR.md file, but it describes no way to verify that a rating is accurate. At version 0.1 it lists four rated projects, and its roadmap ends with asking GitHub and other code hosts to display the rating, so for now its value depends on maintainers choosing to adopt it and rating themselves accurately.