Hi,
At the conference I've also noticed one side effect of LLMs on community
building. I've discovered a new team of developers creating a new
project around one of our historic project, sometimes working around its
behavior instead of trying to improve it at the source, and living in
their own bubble with little incentive to participate to the main
project since their oracle apparently saves them from human
interactions. That kind of situation of course could have happened
before, but use of LLM technologies is likely to favor that type of
community split.
There's some parallel to draw with what happened with Tailwind
(https://www.businessinsider.com/tailwind-engineer-layoffs-ai-github-2026-1)
that has been cut off from most of its users because of LLM interposition.
Even
Le 02/07/2026 à 21:32, Jeroen Ticheler via Discuss a écrit :
Dear all,
At FOSS4G Europe 2026 in Timisoara (Romania) we held a Birds of a Feather
session on the use of AI in OSGeo projects. Around 25 to 30 people joined,
including many core developers who are well respected in our community.
Opinions in the room were strong and differed a lot, which was the point. We
wanted to hear from people who use AI in their work and from people who have
concerns or bad experiences, and would rather block AI contributions completely.
The problem is clear. Some projects now get many AI-generated pull requests
that land on the desks of a few core reviewers. They cannot keep up. A lot of
these PRs are long and give no useful reason for the change. The tone can be
pushy and cold. On top of that, there is little human to human discussion
behind these PRs, which makes the whole process feel inhuman from the start.
When this keeps happening, trust drops, people feel worn down, and conflict
becomes more likely, because they hold different technical and ethical
standards.
People shared their own stories
We heard what it is like to get an anonymous PR with no real person behind it,
where the sender seems to have done little more than press submit.
We also heard positive experiences, where developers have taught AI tools their
own coding style and review habits, then used those tools in two ways. One is
to write new code that is hard to tell apart from your own. The other is to
fight fire with fire: let AI check incoming PR’s and reject the ones that
ignore the project rules, so those never reach a human.
We also talked about newcomers. Many students and young developers learned to
code with AI from the start. Do they know enough to contribute well, and how
will they learn to write the kind of code an OSGeo project needs? The room saw
a link with earlier changes, like the move to new programming languages. Each
time the needed skills changed, and each time people adapted.
A call on tone and community
The first outcome was a call to be careful with the way we talk to each other.
If we lose respect and stop being a welcoming community, the harm lasts longer
than any single PR. *We are in this together.* AI-assisted coding is here to
stay, and it also opens doors, for example for people who do not code but want
to help once good guardrails exist. The room agreed on this and backed it.
A call to build shared AI practices
The second outcome was a call to start a new OSGeo effort where developers
build shared skills, agents, and contribution practices that any project can
use. This is not about copying one person’s style. We have not yet decided if
this should sit at OSGeo level or inside each project.
If you code with AI or review PRs, we want you involved and I invite you to
start the conversation here on our discuss list or on our Discourse General
channel.
Thanks for reading!
Jeroen Ticheler
President of the board of directors of OSGeo
_______________________________________________
Discuss mailing list
[email protected]
https://lists.osgeo.org/mailman/listinfo/discuss
--
http://www.spatialys.com
My software is free, but my time generally not.
LLMs contribute to global warming and brain rot
_______________________________________________
Discuss mailing list
[email protected]
https://lists.osgeo.org/mailman/listinfo/discuss
