[QGIS-Developer] Two proposals -- moratorium on LLM use, and requiring a "proof of life" for contributors

[Dror Bogin]

I'm joining this discussion late, but I wanted to give the viewpoint of a
developer that has tried contributing in the past and has currently failed
to do so.

Going back to Nyall's original points, I can respect #1, but the phrasing
of it (and the further stiffening that Even suggested) will make the
initial contribution process very hard.
I am not a C++ developer by training, but when I wanted to make my first
contribution I used an LLM to look for the right places to make it, explain
what is being done and suggest the edits.
While my PR was not accepted (because of my inability to find the time to
understand, change the code as suggested by the reviewer and get those
changes to actually build and work as planned), I don't believe I could
have even opened the PR or understood the code without the help of an LLM.
I believe that limiting new (and\or existing) contributors to using an LLM
for searching the code, the web and as an autocomplete, should still be an
acceptable compromise for this.

As for #2, I think a financial contribution as a condition can also be too
harsh for some developers who cannot afford it.
Maybe a face-to-face (virtual or physical) with a trusted contributor to
one of the QGIS repositories would be an easier alternative for those who
can't afford the financial contribution.


On Wed, 16 Sept 2026 at 15:19, Laurențiu Nicola via QGIS-Developer <
qgis-developer at lists.osgeo.org> wrote:

> I think contributors using AI are of two kinds:
>
>   - well-intentioned users who run into a bug and use an LLM to fix it
>   - people who instruct their agents "go look on GitHub for some easy bugs
> and fix them"
>
> You want to get rid of both, but the former will manage to file a PR on
> Codeberg just as well — they've been bothered by that bug for a long time,
> their shiny tool managed to fix it, and it's only natural to share it
> upstream.
>
> As for the second category, they're trawling through GitHub because GitHub
> is so massively popular. Imagine a world where gitlab.com would be a nice
> alternative to GitHub instead of a bloated mess with even more AI features.
> They could just say "go file some PRs on GitHub or GitLab" and all would be
> well. If Codeberg had a 25% market share, they would certainly include it.
> And if some project asked you to use git-send-email, the agent would be
> perfectly able to do so (it would probably keep me away, so not all is
> lost).
>
> Of course, that would dilute their "profile page", but hey, I can
> vibe-code a cross-forge profile page in an afternoon, wouldn't that be
> pretty cool? :-)
>
> Laurentiu
>
> On Wed, Sep 16, 2026, at 14:58, Greg Troxel via QGIS-Developer wrote:
> > Laurențiu Nicola via QGIS-Developer <qgis-developer at lists.osgeo.org>
> > writes:
> >
> >> On Wed, Sep 16, 2026, at 14:31, Greg Troxel via QGIS-Developer wrote:
> >>
> >>>   It's interesting that what I'm guessing is large numbers of
> >>>   LLM-generated PRs and/or LLM text in issues is at least mostly in
> >>>   github.  If so, that is a clue that perhaps github itself is a
> problem
> >>
> >> If I may: https://xkcd.com/1138/
> >
> > While that's funny (and everyone who confuses correlation and causation
> > eventually dies), I think there's something to the "github problematic"
> > theory.
> >
> > Tradtionally, projects were centered on mailinglists.  That built
> > commmunity.  With github, people show up and file and issue or PR and
> > don't really join.  github computes scores/profiles and pushes
> > themselves as something that matters, and they push AI.  A project on
> > github that doesn't want AI, and doesn't want to suffer from people
> > collecting gamified points, is swimming upstream.
> >
> > I'm involved in another (very small) project that is list centered, and
> > now has repos on codeberg.  But, propsoed changes are "git send-email"
> > to the list, not PR/MR.   I have not noticed any LLM intrusions.
> > _______________________________________________
> > QGIS-Developer mailing list
> > QGIS-Developer at lists.osgeo.org
> > List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> > Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> _______________________________________________
> QGIS-Developer mailing list
> QGIS-Developer at lists.osgeo.org
> List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer
> Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20260917/f57dfde5/attachment-0001.htm>