With all the discussion around detecting when a code repo contains commits authored by an LLM, I think it is important to note commits like the following in Mozilla Firefox from 2 weeks ago:
-
@jbowen@mast.hpc.social @sarahjamielewis@mastodon.social we need a new browser, its kinda ovbious, both the major browsers are horrible companies now. We need a new browser, we need it to be big, and paid devs to work on it. Volenteer based projects are cool, but there needs to be money behind if
@skymtf @sarahjamielewis @jbowen
Have you heard of https://servo.org/ ? -
With all the discussion around detecting when a code repo contains commits authored by an LLM, I think it is important to note commits like the following in Mozilla Firefox from 2 weeks ago:
"Bug 2011195 - When an agent commits, don't add itself as author"
https://github.com/mozilla-firefox/firefox/commit/71cc24b6a400dbd434e4df37087960d94b764791
I don't think it's a good thing that Mozilla seem to be explicitly encouraging unattributed LLM code in Firefox.
@sarahjamielewis As a tester I find bug reports with titles like "When an agent commits, add itself as author" (#219842) and "When an agent commits, don't add itself as author" (#2011195) highly confusing as they read like they are describing intended behaviour. But a bug report should describe unintended behaviour like "Agentic commits are not attributed to the agent". So before jumping into conclusions about the intent of the authors, I would try to clear up the actual requirements.
-
With all the discussion around detecting when a code repo contains commits authored by an LLM, I think it is important to note commits like the following in Mozilla Firefox from 2 weeks ago:
"Bug 2011195 - When an agent commits, don't add itself as author"
https://github.com/mozilla-firefox/firefox/commit/71cc24b6a400dbd434e4df37087960d94b764791
I don't think it's a good thing that Mozilla seem to be explicitly encouraging unattributed LLM code in Firefox.
@sarahjamielewis I did read a bit further and found that it is because of the dev laziness.
They are too lazy to write commit messages themselves. So they ask the LLM to write them.
And then they are also too lazy to copy paste the generated text themselves, so ask claude to do that.
And they are too lazy to push the commit button, so claude has to push it for them.
And this change is to hide all this laziness. -
With all the discussion around detecting when a code repo contains commits authored by an LLM, I think it is important to note commits like the following in Mozilla Firefox from 2 weeks ago:
"Bug 2011195 - When an agent commits, don't add itself as author"
https://github.com/mozilla-firefox/firefox/commit/71cc24b6a400dbd434e4df37087960d94b764791
I don't think it's a good thing that Mozilla seem to be explicitly encouraging unattributed LLM code in Firefox.
@sarahjamielewis Mozilla is so confident that Firefox users love "AI" slop in their browser that they feel the need to obfuscate its inclusion from them. "We don't need them to know, because we already know that they like it and consent to it! We know better!"
-
@sarahjamielewis
Yet another reason to abandon Firefox and Mozilla in general.Edit: Use NetSurf or Lynx. Or just curl or wget.
-
With all the discussion around detecting when a code repo contains commits authored by an LLM, I think it is important to note commits like the following in Mozilla Firefox from 2 weeks ago:
"Bug 2011195 - When an agent commits, don't add itself as author"
https://github.com/mozilla-firefox/firefox/commit/71cc24b6a400dbd434e4df37087960d94b764791
I don't think it's a good thing that Mozilla seem to be explicitly encouraging unattributed LLM code in Firefox.
@sarahjamielewis its annoying, but i think this is a race simple ideas like this just wont win. The major problems with LLM coding tools revolve around code theft (in training data) and erosion of human skills.. Neither of which you can tell from thier outputs. By far the worst part of AI in tech is the corruption and impunity that pushes it. It is possible without that. You cant tell from the outputs any more than you can tell the political party of a developer in a patch file.
-
@sarahjamielewis Good that they’re not pretending a machine is a person. Bad that they’re not coming up with a new git footer to document their bad decisions.
@mallory @sarahjamielewis willfully bad people will just ignore it
-
With all the discussion around detecting when a code repo contains commits authored by an LLM, I think it is important to note commits like the following in Mozilla Firefox from 2 weeks ago:
"Bug 2011195 - When an agent commits, don't add itself as author"
https://github.com/mozilla-firefox/firefox/commit/71cc24b6a400dbd434e4df37087960d94b764791
I don't think it's a good thing that Mozilla seem to be explicitly encouraging unattributed LLM code in Firefox.
@sarahjamielewis this is the correct approach, if you supervise Ai to build a feature, it's on you to review the code and be responsible if stuff breaks. It would be useless to attribute everything to one Ai account and just be able to finger point at it.
-
@sarahjamielewis I did read a bit further and found that it is because of the dev laziness.
They are too lazy to write commit messages themselves. So they ask the LLM to write them.
And then they are also too lazy to copy paste the generated text themselves, so ask claude to do that.
And they are too lazy to push the commit button, so claude has to push it for them.
And this change is to hide all this laziness.@gunstick @sarahjamielewis Also, nothing like guardrails that can be bypassed by saying "Just do it".
-
@sarahjamielewis this is the correct approach, if you supervise Ai to build a feature, it's on you to review the code and be responsible if stuff breaks. It would be useless to attribute everything to one Ai account and just be able to finger point at it.
@razze @sarahjamielewis The real solution, of course, is not to use "AI" in the first place
-
@sarahjamielewis They even have a full bugzilla section:
@pinskia
For FF that's not an "even". A PR must be preceded with a Bugzilla post, or it will autoclose. They manage issues there, not on GitHub. It's still frustrating, tho.
