Gentoo Linux Plans Migration from GitHub Over 'Attempts to Force Copilot Usage for Our Repositories' (gentoo.org) 37
Gentoo Linux posted its 2025 project retrospective this week. Some interesting details:
Mostly because of the continuous attempts to force Copilot usage for our repositories, Gentoo currently considers and plans the migration of our repository mirrors and pull request contributions to Codeberg. Codeberg is a site based on Forgejo, maintained by a non-profit organization, and located in Berlin, Germany. Gentoo continues to host its own primary git, bugs, etc infrastructure and has no plans to change that...
We now publish weekly Gentoo images for Windows Subsystem for Linux (WSL), based on the amd64 stages, see our mirrors. While these images are not present in the Microsoft store yet, that's something we intend to fix soon...
Given the unfortunate fracturing of the GnuPG / OpenPGP / LibrePGP ecosystem due to competing standards, we now provide an alternatives mechanism to choose the system gpg provider and ease compatibility testing...
We have added a bootstrap path for Rust from C++ using Mutabah's Rust compiler mrustc, which alleviates the need for pre-built binaries and makes it significantly easier to support more configurations. Similarly, Ada and D support in gcc now have clean bootstrap paths, which makes enabling these in the compiler as easy as switching the useflags on gcc and running emerge.
Other interesting statistics for the year:
We now publish weekly Gentoo images for Windows Subsystem for Linux (WSL), based on the amd64 stages, see our mirrors. While these images are not present in the Microsoft store yet, that's something we intend to fix soon...
Given the unfortunate fracturing of the GnuPG / OpenPGP / LibrePGP ecosystem due to competing standards, we now provide an alternatives mechanism to choose the system gpg provider and ease compatibility testing...
We have added a bootstrap path for Rust from C++ using Mutabah's Rust compiler mrustc, which alleviates the need for pre-built binaries and makes it significantly easier to support more configurations. Similarly, Ada and D support in gcc now have clean bootstrap paths, which makes enabling these in the compiler as easy as switching the useflags on gcc and running emerge.
Other interesting statistics for the year:
- Gentoo currently consists of 31,663 ebuilds for 19,174 different packages.
- For amd64 (x86-64), there are 89 GBytes of binary packages available on the mirrors.
- Gentoo each week builds 154 distinct installation stages for different processor architectures and system configurations, with an overwhelming part of these fully up-to-date.
- The number of commits to the main ::gentoo repository has remained at an overall high level in 2025, with a slight decrease from 123,942 to 112,927.
- The number of commits by external contributors was 9,396, now across 377 unique external authors.
Thanks to long-time Slashdot reader Heraklit for sharing the 2025 retrospective.
Context: Gentoo prohibits LLM-assisted contrbtions (Score:5, Informative)
It is expressly forbidden to contribute to Gentoo any content that has been created with the assistance of Natural Language Processing artificial intelligence tools. This motion can be revisited, should a case been made over such a tool that does not pose copyright, ethical and quality concerns. https://wiki.gentoo.org/wiki/P... [gentoo.org]
Re:Context: Gentoo prohibits LLM-assisted contrbti (Score:5, Informative)
The hell are you talking about? Do you even know what natural language processing is? It's a specific way of parsing a prompt that allows for unstructured language use. i.e.: No keywords, no formal syntax, enormous general dictionary. You don't use booleans. You don't use other logic operands. You just say it as you would to a person and it "understands" (spoiler alert: LLMs don't). Natural language prompting has been a unicorn/white whale/Holy Grail for decades, just like AGI.
Now that we've had our CS 101 lesson, in practice they're talking about saying "Write me [x] program" and AI spits out some boilerplate garbage like a bad coder grabbing libraries incoherently from everywhere and adding no cohesive structure of its own. AI will just do randomly seeded, but algorithmically organized, copy/paste. If we start committing libraries of that shit to a code base, eventually the predictive language soup will collapse under its own weight, because it is randomly seeded. There's not going to be a through line. No intent behind it. No Linus bitching anyone out for the garbage code it's writing.
It'll look like what Windows has wound up with with sxs, except there will be no way to figure out how to organize side-by-side in the first place.
You think uncommented code is bad? Wait until you see that shitshow. Gentoo is dead on here. This is a code stability and structure issue. You use AI enough, and you will have neither.
Re: (Score:2)
You underestimate what's possible via the chat window in a modern IDE. I didn't get it either until I joined a team that was using a new-to-me language, and started asking it "how do I..." questions, and it started proposing changes for me. And those changes looked right, worked, and passed code review with only minor revisions.
Of course you have to scrutinize everything it does, and of course it often takes a couple/few tries, and it has limitations... but I'm beginning to understand why people are so exci
Re: (Score:2)
> That's quite the diversion of the topic, but ok.
The person you're flaming replied to claimed that tab completion is AI, and not merely AI, but "natural language processing".
Perhaps you can read the context next time rather than trying to post an off-topic rant to justify your shitty toys merely because someone said something that vibed negative about them?
Or... if you really think it was a diversion of the topic, explain how the fuck tab completion counts as "natural language processing".
Re: (Score:2)
No, tab completion is not an example of natural language processing.
Re: (Score:3)
Re: (Score:2)
Smart move.
Language license shift (Score:2)
Is is possible that someone will create a programming language which has a license
- That prohibits code generation by a LLM
- LLM generated code by use of the standard built in libraries is subject to a licensing fee per character of code generated.
- Training a LLM on that language's code requires a (large) licensing fee paid to the language's copyright holder
And given how Stack Overflow's decade of question and answers will be largely frozen in time for existing languages, there will be little code to train
Re: (Score:1)
It's owned by Microsoft and Copilot actually means "Copilot AI".
It's sad how little the of the old tech ./ remains.
Re: (Score:2)
It's owned by Microsoft and Copilot actually means "Copilot AI".
OK. So what? That doesn't answer the question. How can Microsoft "force Copilot usage for our repositories"?
For example, automated AI "review" of pull requests, making all sorts of nonsensical suggestions.
Re: (Score:2)
Tell it no or just ignore it?
Re: (Score:3)
Tell it no or just ignore it?
You can't globally switch it off for a repo. And short of telling all our users preemptively to ignore it...
Re:What does that even mean? (Score:5, Insightful)
Pretty simple really. You integrate it irrevocably into the commit process. Make a pull request? Copilot is going to have a look at it. Maybe even modify it. Put it in the toolchain anywhere you like, give no one an option to opt out. Done.
Which, you know, is just another way to steal training data to put coders out of work (it won't) and a way to fuck up someone's pull request if it actually does anything other than store it and analyze it. LLMs are all about theft of labor and then not even crediting it, because then someone can track down their own work and demand payment or compliance with GPL or whatever license. Tech bros are forcing it on us because they need free training data. They're not willing to pay for it because that might affect their bottom line. Anytime you use one of these LLMs, you should be getting paid for it. You're providing it prompt data that has a labor cost.
Right on, Gentoo. They can go pound sand.
Re: (Score:2)
Make a pull request? Copilot is going to have a look at it. Maybe even modify it.
Copilot modifies PRs?
Re: (Score:1)
Re:What does that even mean? (Score:5, Insightful)
So, I was going to concentrate on specifics, but then I realized that the main answer is "in ways that none of us could even dream of". Where companies like Google aspire to be truly evil, Microsoft has been doing it for years and has more experience than any other in the field. Firstly, they can just inject issues from copilot directly into your issue tracker: it's worth reading this issue to get some taste [github.com].
If you use a Microsoft product, Microsoft will just install CoPilot into it. In fact there are already desperate petitions begging microsoft to allow Windows and Edge users to opt out [change.org].
If your users sign up for GitHub in order to contribute to your project, Microsoft can default to turning on CoPilot in their Github account.
If you turn off CoPilot for your project yourself, Microsoft controls the project interface and can give all of your users an option to turn it on again and repeatedly prompt them again and again and again and again until they eventually give in and turn CoPilot on over the top of your settings.
If you plan to have children then, as your lord and master, Microsoft can assert their "Droit du seigneur" and inject CoPilot into your children's genetic material.
The possibilities are endless and daunting!
Re: (Score:2)
MSFT will do it the same way they're doing it with their other products. Copilot will become a "feature" that can't be turned off.
You won't have any real choice in the matter.
This is why AI will inevitably consolidate (Score:4, Informative)
Basically whoever controls the major platforms gets to control the training data needed to keep AI functional. Similarly Facebook and Twitter and Reddit are the only platforms large enough to provide enough training data for things like images and text.
This means that we're going to have monopolies for the AI tech of llms. And those basically nothing to stop that.
It makes sense for players like Ubuntu to get out of that ecosystem so that their work isn't being used to help that process of monopolization and consolidation along.
Honestly if it wasn't from Monopoly Linux on the desktop would have happened 20 years ago when netbooks took off. If we have proper antitrust law enforcement then when that guy from Asus or Acer I forget which did that drunken rant about how Microsoft shut down their Linux netbooks there would have been an investigation and Microsoft would have been broken up then and there.
But voters have different priorities than antitrust law enforcement so here we are.
Why no own infrastructure? (Score:3)
git performs best when compiles with -funroll-loops.
Re: (Score:2)
Because git still doesn't have proper pull request support.
Re: (Score:2)
https://git-scm.com/docs/git-r... [git-scm.com]
What took Gentoo so long? (Score:5, Insightful)
Microslop was and still is an enemy of every free operating system, no matter how much chalk their representatives use when they speak. Their "Linux is cancer"-adverts from back then were not just a glitch, they pretty much sum up how Microslop sees Linux.
That "Copilot" force feeding is just another cherry on their existing evil cake.
Re: (Score:2)
Agreed. I moved my repos to three mirrors: A self-hosted Forgejo instance, codeberg.org and salsa.debian.org.
Re: (Score:2)
This is the way.
Re: (Score:2)
Resistance is futile (Score:2)
Slop (Score:2)
Microslop gonna Microslop. I'm never calling them anything other than Microslop even again (unless Nadella pulls his head out of ass, but it's so far up he can see out his throat, not holding my breath).
Re: (Score:2)
I would agree, but they may find something even worse to force on users ...