This reminds me a curl author story who faced issues reviewing hackerone
reports made with various tools:
https://daniel.haxx.se/blog/2024/01/02/the-i-in-llm-stands-for-intelligence/
Since Apache PLC4X is far, far less popular than curl we have much less
visibility and exposure to such issues, but its there.
I am not quite sure how to handle this. Mailing list is sort of filter
which will cut out a lot of contributors which may not be familiar with
this way of communications. Github discussions is same risk as PR, cause
it can always be coordinated by the tool. :/ Slack may be easier for new
contributors as it feels more modern.
Per say we do not need to fight typo fixes, but if someone starts doing
it, then lets pull it further, so one PR will address multiple mistakes.
Hopefully we will be done with it for a while.. or until another tool
will decide to re-phrase sentences we made in docs.
Cheers,
Łukasz
On 10/7/25 17:32, Christofer Dutz wrote:
I this case it appears to have been „someone“ admitting to have used AI to
draft the PRs. Not sure if he wanted to pimp his resume, or if „he“’s a bot or
has malicious intent … seems like communication on the mailings was too much to
ask for. At least he said he subscribed … so he should be getting this email
too ;-)
If we’re doing him wrong, here would be the place to speak up, but I doubt
that’s gonna happen. If it does, I appologize but would really like to learn
more about the motivations.
Also … perhaps we should use AI ourselves so spot things AIs could whip up PRs
… the fewer of these that we have, the less AI pre we have to handle ;-)(
Chris
Von: Lukas <[email protected]>
Datum: Dienstag, 7. Oktober 2025 um 15:08
An: [email protected] <[email protected]>
Betreff: Re: [DISCUSS] Do we want to make communication a requirement before
merging PRs?
Yes an additional interaction via mailing list, where the PR is explained helps
us to at least make the probability higher for an actual person and not a bot
generating code.
Am 7. Oktober 2025 14:23:13 MESZ schrieb Christofer Dutz
<[email protected]>:
Hi all,
Afer we jus got another set of AI PRs, I think we should discuss how to
generally handle these types of PRs.
They usually handle super trivial stuff where merging is sort of a no-brainer.
However, merging a PR gives access to some GitHub Runner resources. It also
lets „AI contributors“ appear more credible in other projects (if you don’t
actually look at the PRs that got merged).
I personally would like us to live the Apache moto: "Community over Code“.
What do you think? Should we make communicating on the mailing list before
having PRs merged mandatory? If the submitter declines or disappears, we simply
replicate the changes and close that door for automated PRs.
What do you think? I would like to add this requirement. Email … slack … up to
them … I just want some form of interaction that helps us at least raise the
bar for AI bots.
Chris
