Attribution

Most tools let you claim how much a model helped. Byline measures it, and the claim a journal, a regulator or a publisher asks for is produced from the version history rather than from memory.

The chain, in four ordinary git steps

1. Snapshot before. Uncommitted edits are saved before the model touches anything, so its turn starts from a known point.
2. Commit the turn. Its changes are committed with the model that produced them named in the message.
3. Replay ownership. The history is replayed word by word to work out who last wrote each one. The editor tints accordingly, so it is visible while you write.
4. Disclose at export. The AI-use statement is generated from that count, with a regression test guarding the numbers.

Why it is plain git

None of this is a private format. Clone the repository, run git log, and the same record is there without Byline and without MAH7. A disclosure only one vendor can produce is not evidence of much.

What it does not claim

It measures words, not thought. It cannot tell whether the sentence a model wrote was the sentence you meant, and it makes no judgement about how much assistance is appropriate in your field. It answers how much and where, which is the part that was previously impossible to answer honestly after the fact.