I'm super impressed with the reviews from @coderabbitai -- they're an order of magnitude better than Claude (and two orders of magnitude better than copilot) even with the latest models. And free for open source projects!
@coderabbitai how do I refer to you in a commit? e.g. "Co-Authored-By: coderabbitai" perhaps? Claude uses "Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>" if that helps.
@hughsie please don't add these trailers to the commit log. All they do is free advertising in the place that cannot be changed for the rest of the project; and they don't help anybody with reproducibility or historical information, because you cannot reproduce the results even if you, by chance, had access to the model state at the time. If you want to make it clear that you used code assistance, put it in the pull request description.
@ebassi doesn't that seem a bit disingenuous? I didn't write the patch -- I just asked the agent to do something on my behalf. I'm not even sure if I'm the *author* :/
@hughsie if you didn't even review the patch, and you didn't validate it, then why are you committing it? If you did review the patch, then you're the author, not the agent. The agent is not a person, and cannot claim authorship.
@hughsie @coderabbitai https://docs.kernel.org/process/coding-assistants.html#attribution I would go with this
@ebassi I most certainly reviewed it, and more than probably tweaked the code style and changed the commit message to make more sense.
@swick @coderabbitai yes, Assisted-by seems more correct than Co-Authored-By -- I'll try to get Claude to do that in the future.
@hughsie there you have it, then: you are the author.
Disclosure is meant to help maintainers review your contribution; if this is your project then you should know what you did, and the commit log isn't really going to help you. Unless you want to go through the log in 5 years time, and you need to find a commit that was code assisted and you need to revert specifically for that reason.
@hughsie @ebassi if you reveal that you are using AI to generate code in the private sector, you undermine your value as an employee, which will either prevent you from receiving pay raise and or eventually get you fired and simply be replaced by that same AI you yourself used to generate the code you submitted.
@hughsie @ebassi you are an good example why all the value resides in "senior developers" when AI is put to practical use as an tool.
You are an individual that understand the code thus can differentiate bad AI results/suggestions from good AI results/suggestions.
What the above means is that Uni grads will have an extremely difficult time finding jobs in the future, in America and Europe due to lack of experience and coherent government policies how, when and where AI should be used.
"And free for open source projects!"
That's nice, but it used to be...actually F/OSS. Then they closed the source so they can go make lots of money.
I'm more interested in https://github.com/The-PR-Agent/pr-agent . That's a similar story - Qodo also started out open source then closed up shop - but they at least are allowing 'The Community' to actively develop the original F/OSS codebase from the point they forked it.
I need to find time to write a Forgejo backend for it...
@hughsie @coderabbitai old coderabbit repo, now archived: https://github.com/coderabbitai/ai-pr-reviewer . you can rewind a few commits to get rid of the 'closing up shop' ones and see what it looked like. I don't know if there are any actively-developed F/OSS forks.