UPDATE: It was the inline suggestions. If you use it to fill at least a character or word it will decide that it owns your code now.

I accepted a typo fix in a changelog while correcting it. Appears as tab to autocomplete for those unaware. Still seems like a reach.

This only happens on commits where autocomplete (via Copilot) was used, which I think is on by default.

Still that doesn't seem very reasonable, LLM autocomplete seems like a basic editor feature nowadays.

I have a rule for Claude to stop watermarking the commits. I feel exposed when it's doing it and I rewrote the entire Claude assisted commit history after switching from Copilot to Claude.

The problem isn’t “AI in the workflow.” It’s AI claiming authorship by default.

If I didn’t explicitly use Copilot for that commit, adding it as co-author is not assistance, it’s misattribution.

Friendly reminder that all (I sure hope it does all...) the Copilot kerfuffle can be disabled in `chat.disableAIFeatures` flag in the settings.

How does this look like?

Good times for being a Vim guy...