Codeberg, are you serious?!? transcript

As of recent, I've been pretty harsh on GitHub, and I think it's for good reason. The platform kind of feels like it's falling apart. It's regularly the biggest bottleneck in my workflows, and I feel like I spend more time waiting on PRs loading than I do on actually writing the code or having my AI do it in the first place. That's why I recently did a video about what we could use instead, and I settled on Codeberg, because I was really impressed with it.

It performed better than expected. It was genuinely a really pleasant experience. It's what a lot of projects have been moving to recently, even ones I'm iffy on like Zig. And I ended up donating a bunch of money to Codeberg and encouraging others to do the same, because I really wanted a proper open alternative to GitHub to be able to be out there and successful, especially after I heard about the drama with the Codeberg and Forgejo guys and Gitea and the way they split out.

It really seemed like a project that was well aligned with what I wanted, and I was more than happy to support them, shout them out massively, donate money, and send as many of you guys as I could to go help however you can. But today I saw this post. "Proposal Assembly 2026. Terms of Use Extension to Prohibit LLM Extrusions." And I was immediately scared, but assumed it can't be that bad. Maybe they're just trying to ban contributions to Codeberg itself using AI.

Not my favorite thing, but could be reasonable. Then I saw this poll that they ran internally to disallow vibe-coded projects entirely via the terms of use, which got a majority agreeing vote. I can't believe I just finished filming a video about Linus's entirely reasonable takes about AI being a tool that can benefit and hurt maintainers, so we should try to make it more helpful, just to be linked this horrible announcement from the place I like hosting Git repos most.

This video's going to be a little different. It's both going to be me crashing out about how misguided this decision is, as well as a plea to Codeberg to reconsider, and hopefully some reasonable arguments as to why. Either way, I need to make some donation money back, so we'll do a quick break for today's sponsor. If you've watched me for a while, you're probably already aware that I love building new things. There's a problem though, shipping them isn't the easiest and as cool as things like Lakebed are, they're still a problem when I'm trying to get these apps to users.

If I want to charge, if I want them to be able to sign in across different platforms, if I want them to manage their off and users and orgs and all of that themselves, I run into a lot of issues unless I'm using Clerk, then they all go away. Clerk has made my life so much easier for building real apps that face users, in particular because they have off all figured out obviously, but more importantly, they have payments figured out too.

And their off goes way further than basic front and back end web stuff, by the way. Their mobile SDKs are unbelievable. That's why we're using them for all of T3 Code. When I'm trying to build an app for my users, I just want to build the functionality. I don't want to build all of these account management flows and mobile off layers and billing platforms. I just want to go ship the thing. And Clerk has made it so much easier to get there.

The combination of their component libraries, their best in class SDK that is written for both humans and for agents, and their billing platform as well has made it so easy to integrate. If you don't believe me, go look for yourself. The T3 Code repo is heavily using Clerk and you can see a great implementation there. To be frank though, there's not much to see because it's so little code. Get off and billing right at soydev.link/clerk.

I'll start with this pull request that was filed to make the amendment. There is an official thread in the forums where they went back and forth about this, but sadly, I do not have access. Cool. I'm not allowed in because I'm not a member of the Codeberg official organization. So, they're linking this and not sharing it. I will say first and foremost, as a fan of open discussion for open projects, I don't love that the discussion is behind a restricted wall that cannot be seen.

I do appreciate their desire to keep their forum clean and outside of all the types of like rude comments and speculation that they're going to get and I'm sorry in advance for any I add on here. On that note actually really really quick this is important. Do not harass anyone on the Forgejo or Codeberg teams. If I see any of you guys giving any of these maintainers any [ __ ] or making their lives harder as a result of personal disagreements, you are not associated with me.

I take Do just don't do that. It does not help anyone at all. I do not wish any ill will to anyone involved here. I am just concerned about the direction and I'm hoping that with a little bit of back and forth we can steer them in a better way. And I'm assuming that their desire to keep the forum private is part of their attempts to prevent harassment of the maintainers. I respect that and I will do my best to keep you guys from harassing them.

If anyone does and they're a part of my community, let me know so I can kick them the [ __ ] out. I did find this thread where they discuss their stances a bit as well as the complaints about the private discussions come up in here too. I think it shows just how far gone these folks are. Codeberg should establish a project and platform wide policy against use of AI to generate code, text, or media. A platform wide policy.

It's clear at this point that the technology is a detriment to the long-term maintainability and quality of software projects while otherwise raising many ethical concerns being developed by corporations that back war crimes from data scraped without consent at the expense of website operators and to the detriment of the environment then serving to bury the web under mountains of soulless generated content, dilute audiences, and de-skill workers while attempting to devalue genuine labor.

Most people are into open source because they care a lot about software. Occasionally you find one who's here cuz they're unemployed. I'll leave it up to the audience to decide which this person is. They have previously taken a strict stance against cryptocurrency projects. Are they still on the platform? They are. So they are complaining about these cryptocurrencies, but they're all still on on site. They're all still getting updates.

The mere nature of working on free and open source software is political. This does not imply that Codeberg supports political parties. Oh boy. Apparently, this vote only happened yesterday and it passed. So, they just decided they're going to ban cryptocurrency and they're going to ban all AI-generated projects. I I'm going to fixate on this first bit a little too much cuz it's really frustrating for me. In particular, this call out that the technology is a detriment to the long-term maintainability and quality of software projects.

I'll take the hard stance here. Anyone who believes this writes worse code than an AI model from a year and a half ago. If you legitimately believe that top-tier AI tools are actively hurting maintainability of projects and not helping at all, that's because your brain is slower than [ __ ] Sonic 4. And there's nothing we can do about that. You've already lost the plot. And if you want to take it up with somebody, feel free to take it up with Linus, who said verbatim, "There are other questions around AI, like what is this going to pan out to economically in the end, but the is it useful statement is no longer one of those actual questions.

Anyone who doubts that clearly hasn't actually used it." Correct. I hate that I have to cite Linus against Codeberg here, but I do because their stance is anti-progress of software. It is guaranteeing that software on Codeberg will be slower, worse, and less secure. Because if I can't patch software that is hosted on Codeberg using AI, I can still find exploits in software on Codeberg using AI, and I can use those exploits outside of Codeberg to break all the software on the platform.

I'm not saying you should, in fact, I'm actively saying you shouldn't because there's no longer any safeguards to prevent that otherwise. The best thing you can do if you want to keep your project safe is move it somewhere where you can use AI to actually verify it, check it, keep it safe and update it. Funny enough using open weight models because all of the best models, those being Fable and GPT-5-6, have hard restrictions on how much you can use them for safety stuff.

So, thank you open software for making it possible for us to use AI to verify and validate our software and its safety. But, none of that's going to happen at all in the land of Codeberg, the land that I donated thousands of dollars to, which is not something I'm very jazzed about here. And I'm like, cryptocurrencies number one [ __ ] hater. I hate modern crypto so much. It being in here as well is stupid, and it should not have been banned.

They're talking about how they needed a member of like the official Codeberg team to propose a vote, which our friend here, Jadane, did. The notes of the annual assembly of the association are indeed not public, but the summaries are released on the blog, for example, this one from 2025. Generally, when it comes to shaping the policies of Codeberg, it's probably easiest to become a member, as that way you'd get access to the internal organizational repos where people work together on preparing motions to bring the annual member assembly.

In case there's a perceived blocker, membership in the association isn't limited to Europeans or Germans. Only some elected roles come with particular legal responsibilities, do and those are a little bit due to the way German association law works. I'd like to argue that there's a non-trivial subset of people who are not willing to become a member, but would like to be informed on what policies are being actively discussed and decided upon.

There's a good reason why Codeberg decided to not cater to that group, but I cannot think of one. In any case, I'm grateful Codeberg exists in the first place. Yep, totally agree here. Thank you, Ethan, for speaking for the people. This is a very good point. More questions were asked why are they so opaque? It's virtually all volunteer driven associations. There's orders of magnitude more work than there are volunteers to do the work, and as such, efforts need to be prioritized.

Given the current balance of work and volunteers, it mostly means there's no resources to do extra effort to regularly push out information about discussions beyond the outputs that were voted on. But, starting at $12 a year, you can join to volunteer your time to change that. Okay, so I should be allowed in cuz I donated a shitload of money. You guys better let me in. I'm going to apply after my stream. Somebody asked, "Why are these not held in the first place?" And we got a pile of word vomit from Natus, who I'm going to assume didn't use AI to generate this.

He's a member of the Presidium and he and he supports restricting AI slop. He says restricting instead of banning because finding the red line is a topic to be determined. He responded to the question of why aren't these discussions in the open with, "Why can't you read Stephen King's draft manuscript from a new book? Why can't you read a lawmaker's law the minute they start writing the first sentence in their word processor?" Uh you can't read a new book immediately because it has to be paid for and it's not a free and open source thing.

The point of free and open source is transparency. Citing Stephen King is a questionable choice here. "Why can't you read a lawmaker's law the minute they start writing the first sentence in the word processor? Because they usually talk about what they intend to write and share their policy positions and allow for a discussion before they start writing the law. We can philosophize about radical transparency and direct democracy and where the line is, but depending on who you ask, one might say it's ready when it's ready and nothing is official before they are made official.

And also, it depends on volunteer availability." So, how about you just put this in the public? Have you seen the email list that I was citing earlier where everything is in the public? I have to have a custom Chrome extension just to hide email addresses so I don't accidentally share email addresses in my videos. We can be opaque, but I don't think we're that opaque. We're currently working on improving our transparency in certain key areas such as finances.

Apparently, a motion passed. God, that better not be something that where they're mad at me for donating an existing. Terms of use changes and privacy policy changes are deemed as policies of the nonprofit and they're voted for by the members per the bylaws, many of which say I'll join Codeberg because they dislike Co-pilot, by the way. But even if that if God, that's such a stupid [ __ ] comment. Oh, [ __ ] hell. This is the type of comment where like I'm going to be a [ __ ] here.

I am sorry. If I'm going to get [ __ ] in the comments for anything, it's this. When I see people in and around my spaces leave statements like this in their communication, I look for the quickest way to remove them. Whether that is firing them, distancing myself, whatever. This type of sentence is very much indicative of a person who is not trying to communicate. They are trying to punch. Thankfully, my chat seems to get it.

To be fair TheO, if I donate thousands of dollars to an organization just to cover this type of thing, yeah, I'd be crashing out, too. Yeah, they're doing a [ __ ] cult here. Like this type of person should not be allowed decision-making processes because they're acting out of spite, not out of values. And if this is their attempt at transparency, they're dishonest or lying. They just are. Chatters are saying this doesn't feel as professional as they'd expect it to be.

Yeah, it's not at all. That's probably why they have the [ __ ] forums private because it would be really, really embarrassing. I guarantee in those forums there are many people crashing out at me specifically because they're so upset that I decided to give them money. How dare we associate ourselves with that SlopTuber TheO? How dare he give us money so we can fund our business and we can create free and open-source software?

It's rare that somebody writes this much slop by hand. I'm almost impressed. Anyway, the time explaining this could have been spent into making a good human first proposal. Jesus [ __ ] these people, man. I thought that providing a thoughtful explanation of the way I see these things might be enlightening. Keep in mind that all of these are personal opinions. Thanks for letting us know at the end there. Taking an unapologetically political stance against a fascist on many fronts technology is the right thing to do.

Do you think there are people in those forums that called me a fascist? The guy who donates as much money as they can to open-source? The guy who [ __ ] campaigned for Bernie Sanders in college? They're calling me a fascist because I like that my code can update itself now? You [ __ ] kidding? If any of the maintainers are upset with this and want to leak internal comms to me, I won't publish without permission. My DMs are open on X, the fascist app.

I'm sorry, but there's nothing to discuss here for non-board members. As I've got the feeling that this will get heated and possibly have to moderate the discussion, it's locked till the vote is over and there's a result to share. Good news, they just published the article while I was filming. Protecting our open-source commons from LLMs. In brief, two motions regarding artificial intelligence and large language models were voted on among Codeberg members and passed.

We are promising to not use any of your data to train LLMs and explain what the planned terms of use change is for Vibe Coded projects. We believe that LLMs endanger the free libre software ecosystem as a whole. For those of you who have always been confused about why folks like the GNU/Linux people get so much [ __ ] here you go. The Codeberg annual assembly is the meeting that puts power into the hands of our active members.

Proposals are discussed live and later voted on asynchronously. Since LLMs are an emerging but controversial technology, it is not surprising that two of the votes were concerned with Codeberg's position about this technology. The 14-day voting period ended yesterday and both proposals were accepted. The first vote was a statement about Codeberg EV stance on using your data to train LLMs. This is totally reasonable to promise to your users and to your members that you will not use their data to train LLMs.

Totally okay with that. Absolutely reasonable. The second vote was more controversial but was also accepted with a 358 agreements to 144 disagreement vote with a high voter turnout of around 50% of active members. That implies a change to our terms of use to prohibit Vibe Coded projects. We'll share thoughts about the practical impact at the end of the article. For what it is worth, the change is here if you want to read it.

It is an additional term in the terms of use. You must not share projects that mostly consist of code written by {quote} generative AI tools, including services such as Claude and OpenAI Codex. Such projects have an unclear copyright status, you can see in other requirements, and furthermore have little safeguards to ensure that they do not include harmful code. Okay. I'm going to do a poll with my audience here. What do you think is more likely to be secure?

A random vibe coded package using modern tools like Fable and GPT 56, or a random below average dev's package? Which do you think is more likely to be secure? Genuinely. Put your real answer here. 90% of my chat agrees that a random vibe coded package is more likely to be secure. There's a pretty obvious reason for this. A lot of time and money has been spent making the models generate better than average code. Every single time a new model comes out, they have put more time into trimming the data, refining the model, and making it behave better and better.

Similarly to how a developer gets better after years of writing code and more and more experience being built up, LLMs have teams of people putting billions of dollars of effort into doing the same, but then once that LLM is out, everyone now has that as their floor. To be clear, the models don't get to learn the way a human does. Where if I hire a shitty dev, I can teach them how to be a better dev. If I hire a shitty model, I can't make it a better model.

But the floor has been raised massively. LLMs are so much better at coding than most developers now, especially in a general sense. Many of my chat are saying they would have voted the other way a year ago, and I would have too. But since Opus 4.5, then generally speaking, I've read a lot of code from AI, and I've read a lot of code from people. The average person writes worse code than the average thing using Opus 4.5 or newer.

So again, this idea that it's less likely someone uploads harmful code if they can't vibe code it is absurd, especially because the two tools cited here actively prevent you from writing harmful code. Code X and Claude code both have really strict safeguards preventing you from using them to build malicious software. There is nothing preventing me from doing that with my brain other than the law. I promise you guys, and this is just an honest take for me as a donator to Codeberg to you as the maintainers that seem to just be behind here.

I promise you guys, humans are way more likely to create and upload harmful code that LLM's can at this point in time. That is just obvious and I wish you understood it. If you disagree, my DM's are open, my email's in the bio. I'm down to chat, I'm down to convince y'all. I'll even do a internal presentation if that's what you need. I want to help because your takes are wrong and it's going to cause problems. This is just factually incorrect.

Let's see how they justify it. They open with the title, "We all pay for hungry LLM's." LLM's are a costly technology and this cost keep rising as the companies providing them have to start recouping their investments. They're not only costly for those who use and explicitly subscribe to these services. The costs are not only hidden in normal cloud and service subscriptions that cross-finance these innovative new features that you never asked for.

LLM's are so costly that companies are externalizing the costs on massive scales on those who don't use them in society at large. Increased hardware prices, energy use, and environmental damage. We all pay for it. What are you linking here? The author's complaining about dealing with hyper-aggressive LLM crawlers at scale. I agree that is bad. The outrage that I and others have been seeing from the scrapers from Open AI and Claude and even companies like TikTok recently is egregious.

That should not be a thing. But how the [ __ ] do you think banning LLM generated code from your platform is going to help prevent AI scraping? That is such a dishonest thing to include here, it's insane. They're also going to ban Mac users from using Codeberg because they don't align with free and open source software because their stuff isn't maintainable or repairable enough. Cuz that would be an easier to justify stance.

I genuinely believe that. I'd be more in favor of them banning Apple users than of banning LLMs because there's a much clearer throughline to their beliefs here. They call out the strain servers due to nonsensical crawling. Again, I agree. Ban all the crawlers you want. I will help fund the banning of crawlers. I'm so down. The development team of none. Using LLMs to work with code gives you a kick of adrenaline. You can develop at a rapid pace, build things as if you had a large team.

This is The English for this isn't great and I understand they're not native English speakers, but doing my best to read this here. Feels like you have a large team but you have none. In fact, you're often alone working with a statistical machine that turns energy into code. It seems like many vibe coders don't realize that they don't actually have a community around them. They build projects as if they did and they spend resources accordingly.

We see projects having a lot of code activity, heavy CI and CD testing, frequent and large release binaries. Sometimes it feels like the amount of supported platforms exceeds the amount of actual users. I agree here. There are projects that are way too big for what they are yet. To them it seems ridiculous to see projects with a single dev and virtually no users consuming as much or even more resources than some of the largest community projects on Codeberg, which operate frugal with CI CD and storage resources.

We do not believe it's reasonable for Codeberg to invest their precious donation money into hosting of large ghost projects. Again, absolutely agree here. I wish they opened with this. If they had opened with the expense on our service for hosting these apps is too great, here is the percentage of our CI that we estimate went to people building projects with no users with vibe code. Awesome. I'm in favor of that. We can restrict those users from the free CI.

We could flag in the people we like for CI or even do the ban. I'm okay with that. But just do it with this. If the goal is to actively protect Code Berg and its users, there's a conversation to be had. But notice how far down the article we are before we're talking about actual things that actually cost Code Berg at all. Everything else here is political dribble. It's useless. It's nonsense. It's them caring too much about things they clearly do not understand.

This is a real problem. If this is costing them money, if hosting these projects is making it so others who don't use AI can't Cool. I get it. Awesome. I'll donate more money for this. And if you want to block these things on paying users or confirmed users, cool with all that, too. If that was the conversation, we can have the conversation. But you're not having a conversation, Code Berg team. You are telling me why I'm wrong when you don't understand me.

They then complain about hardware prices. Yeah, it sucks that computers are more expensive. It'd be awesome if there were technologies that we could use to increase the likelihood that there were open alternatives to these expensive things so that we could use those. It'd really be a shame if, since we can't use the computers that are high-end because the LLMs made them cost too much, that we can't use the LLMs to make cheaper computers better or to design our own alternatives.

We're drawing a hard line here between the people who can use these tools and the people who can't. And what you are doing by doing this is massively increasing the gap between the quality of free and open-source software and software that isn't shooting itself in the foot for excruciatingly stupid political reasons. I want free and open-source software to win. You guys want free and open-source software to feel good.

That's the difference. I want open-source software to be as incredible as possible, as high-quality as possible, and just to enable more people access to more awesome [ __ ] What this article does and what this policy change does is the opposite of that. It is restricting open-source software from being able to get as good. They call it a growing digital divide, and they decide they want to put a brick wall between the two sides.

[ __ ] great. They complain about the environment consuming costs of both electricity and drinking water. They're not going to really bring up the [ __ ] water thing, are they? I've been putting off doing that video for a year cuz it's not a real [ __ ] thing. There are certain towns that have really bad water access where if a data center opens, they will have slightly worse water access for a few days when they fill the data center with water, but it's a closed [ __ ] loop.

They get the water once. This has happened to maybe three times in history. And I agree we should have a conversation about the one or two towns that were affected with slightly restricted access to water, but you realize the data centers use about 1/1000 the amount of water that corn farming does in the US, right? And that data centers use less than 1% of residential electricity. It's basically nothing. Hint Greens also covered this if you want to check out his stuff.

It was actually very solid. There is a real problem with energy usage here. I absolutely agree. US electricity generation has barely moved since '99. China energy capacity has massively increased in that same time. We need more electricity generation. We need it bad. And if we're doing it with [ __ ] gas and coal, that is horrible for the environment and will use more water and that is bad, which is why we need to do something better, at the very least nuclear, but ideally more solar farms, more wind farms, more generation methods that aren't horrible for the environment and aren't wasteful.

If you really care about water, I hope that you're advocating against corn subsidies. If you really care about electricity usage, I hope that you're advocating very actively for green energy. And I'll take a stance on this. Right now is one of the most important times in history to be advocating for green energy. Solar and nuclear are great. Solar has its issues because it's only available during the day and energy storage is expensive.

I could go off for a long time about energy stuff right now. That's a really good and important conversation for us to have, not AI bad cuz water. You're doing a disservice to issues I and others care about here because I agree we need to do more for electricity, but we're not. Let's fix that. Let's advocate for that. Let's align and work together to get more green energy in our country. Do you know what's really funny?

It's been measured now. Data centers tend to lower electricity rates for the places around them. And when you compare data center energy usage to things like golf courses, lawns, and crop irrigation, it's a rounding error. It's nothing. You're performatively mad at this. You either haven't looked at the numbers or you have and you don't care because you just want to be a dumbass. It's insane. It's insulting. I I am annoyed.

This is I'll be really transparent here. One of the things that triggers me the most is when people hate on a thing I hate in such an absolutely incompetent way that it degrades the quality of the conversation I can have. Like when people hate on Anthropic for things that don't matter or that they didn't do, it pisses me off because I hate Anthropic and I want to tear him down for the actual [ __ ] they [ __ ] up. When people hate on Apple because they don't understand battery game that was Apple trying to preserve your phone over time, it pisses me off because I can't have as good of conversations about Apple anymore.

I hate when people do this. There's a really important conversation to have about energy generation and it's being taken from us by these dumbasses who are allergic to looking at numbers and it pisses me off. Also, they're complaining about Frankfurt, Germany, which is particularly funny to me because Germany is well known for one particular thing in the energy world. This chart is showing nuclear energy in Germany over time.

They banned it. They went from over 20% of their energy being nuclear to less than 5%. If you're wondering why energy problems are happening in Germany, it's cuz they took down 15% of their grid. And in order to make up for the difference, they had to open up more mines and burn more coal. Actually insane. And they've been bragging about this constantly. They are now at 0% by the way, if you were curious. I see a handful of Europeans making really dumb comments about the energy usage in Germany.

I want to be very clear about this. Their recent push to do more green energy is great and noble and I love that they did it, but when they phased out nuclear, more coal generation and cross-border electricity trade, which was spoiler, mostly oil from Russia, had to be used than otherwise would have been necessary. Sorry. Germany became much more dependent on Russia's energy generation, which was a massive hit to their sovereignty and also to the goals that we hopefully share here, which is greener energy generation from less [ __ ] up places.

They have since spent a lot of energy, time and money to try and get more green energy generation, but they've also had to lower the energy generation across the [ __ ] country. And yes, I used ChatGPT to answer this cuz I wanted to make a [ __ ] point by burning more energy. All of this information is public. If you want it, it is cited here. I'm sure you can find it if you prompt the same thing. And you can see very clearly here there was a big bump in coal usage after these changes occurred.

And their total domestic generation has plummeted from 633 terawatt hours down to 509 in that time. And also of note, they're now a net importer. They have to import energy because they're not generating enough for their own god damn country. So share. Cool. You You did it. You got rid of nuclear. Hope you're proud. Crazy how nuclear energy doesn't appear anywhere in this document. It's almost like the thing you care about isn't the energy usage, it's your performative [ __ ] And now they say collaboration's at danger.

It's not purely the digital and civic infrastructure that are impacted by the use of LLMs. The free and libre software ecosystem, of which we consider Code Week an important part, is a social phenomenon centered on collaboration. Working in this way is only possible thanks to free sharing and mutual learning. This includes even very small tools that are shared and reused around which collaboration can start. In contrast, by adopting LLMs, people tend to code single use software from scratch.

While this leads to an increase in shared code, it's mostly code that not only has not been written by anyone, but it's also not maintained by anyone. Do you know how many free and open source alternatives to things I have built with the help of LLMs in the past few months alone? Here's one, Lawn Video. I did not like Adobe buying Frame.io and did not like where it was going and as a result, I made Lawn Video, which is a free and open source alternative to Frame that is cheaper, self-hostable, easy to do whatever you want to.

And if you really hate the Labs, I got a gift for you, t3.codes. It's a free and open source alternative to using tools that are closed source like the Codex desktop app, the Cloud Code desktop app, or those types of things. And it also supports Open Code, which is a fully open source harness and layer that you could use with Open Weight models. Having a full end-to-end open stack for your AI usage. And I used AI, I used admittedly Frontier Labs to build some of this and I think that's awesome.

But apparently, we're just going to make single use software and that's going to hurt the open source community even though I alone have made multiple successful open source projects thanks to AI. Looks like we might have another point I will kind of agree with, which is losing trust in each other. I have said for a while now, trust is a really, really valuable thing to build. It's one of the most valuable things you can have, especially now, and AI is hurting trust overall.

That is bad. That is scary. Let's see what they have to say about this. Hopefully, we don't hate each other. The widespread use of LLMs in open source software is instead becoming a multi-dimensional attack on the trust between contributors and the very idea of convivial collaboration itself. Maintainers are under an increased workload due to people submitting often well-meaning, low effort LLM generated contributions that require substantial amounts of time to review.

Yes, I agree. This is bad and I will even throw some mud at the labs here. Our friends at Anthropic and OpenAI should have done way more about this early. They should have built more tools, technologies, and solutions to help open source maintainers deal with this and to jump in front of these things as well. At the same time, it's becoming increasingly less clear which projects are maintained by experienced devs and which ones are LLM generated without any meaningful human oversight and input.

Oh, are you saying that it's hard to distinguish between and LLM maintained projects? Is that cuz the quality is hard to distinguish between? Huh. You do know that there has been a recent surge of open source projects where the maintainers got burnt out, handed to somebody else, and then the human used that to exploit things, right? Just just saying, in case you didn't hear about that. If LLMs can be used to make maintainers' lives easier, so they're less likely to have to hand off the project, I think that's a good thing.

Great summary from chat here. Thank you, ABCD. I want to hate it, but I can't tell what to hate. Yeah. In the case of copyleft projects, LLMs additionally will lead to license laundering, where copyleft code is stripped of its reciprocal requirements by generating it out of the training data. If you have proof that there is licensed code that's in the training data that should not be, and the model can regurgitate it without internet access, you have a really compelling lawsuit and I genuinely encourage you to make that.

We observe an increasing trend of mistrusting each other up to the point where people who put in actual effort to analyze issues and share their suggestions are being accused of having used LLMs when they did not. Oh, are you saying that effective people and LLMs look a lot alike? Are you saying that maybe being accused of using LLMs feels bad and people don't like that? Are you saying that you guys are encouraging this type of thinking and work so heavily that people are being [ __ ] as a result?

Weird. Crazy. At the same time, others are instructing their LLMs to hide their traces and actively avoid common patterns, prompting others in reviewing contributions and communications more carefully for signs of machine generation. Oh, are you saying that these people for some reason wants to hide the fact that LLMs helped with their code? I I wonder what that reason might be. I wonder if it has anything to do with everything else you [ __ ] have said in this article.

Entering a vicious cycle. Together, these forces make collaboration not only harder but less rewarding. With the transaction cost of collaborating increasing, people are becoming less likely to contribute to creating high-quality software projects and more likely to vibe code a one-off software that is specific to your needs and won't evolve beyond. I disagree. I think that the cost of collaboration going up is indeed bad, but you're framing it incorrectly here.

Although often well-intentioned, sharing the result of a prompt and calling it Libre software does not make the world a better place. Codeberg is not and does not want to be a place to dump such generated single-use software that no one else will ever look at. We are a place for people to collaborate and improve software together. Within this context, the recent votes could be understood as a reconfirmation of those principles.

I hate this. We acknowledge that many devs have started to embrace LLMs as a tool in their workflows. Some use it extensively and rarely code by hand. Others delegate only specific tasks to it. We understand that you want to know how the change affects your projects going forward. While we can't give an easy answer, we'll share some remarks that should address most of the concerns raised in the discussion. If your work fits in these cases, it's unlikely you're affected at all.

Projects that have an active community that cares about and maintains the software. Projects with a significant pre-LLM history, and maintainers who unknowingly or willingly accept LLM-generated contributions from other contributors. If your project otherwise does not involve the heavy use of LLMs. We will also not spend significant amounts of time and resources to automatically scan content on Codeberg. So, while the following use cases are discouraged, they are likely to be tolerated in practice.

Things like side projects and experiments with little resource usage, and specific tools and custom scripts that would be unlikely to find a community anyway, even if they were not LLM generated. However, we also need to be honest about certain use cases that may no longer be welcome on Codeberg. If you see yourself on the list, you don't need to move right away, but there may be other places that better fit your needs.

They list alternatives to Codeberg. These are some of the other Forgejo instances they recommend. Disroot, self-hosting, a list of other things, and then SourceHut, which is an alternative entirely. Cool. Great. Thanks, really helpful. So, the things they are banning are projects that are created by LLM agents in autonomous ways, projects written and maintained with heavy use of LLMs, projects where the amount of resources is significantly larger than what the involved amount of people could have created by hand.

This, absolutely ban. I'm totally cool with that. The rest is garbage. Projects heavily tied to the LLM ecosystem, like LLM written tools to ease LLM usage. So, T3 Code will never be on Codeberg. Great. And users sending LLM contributions in violation of project's custom policies. Disgusting. I hate this. This is so pathetic. I honestly expected better. I Every interaction I had with the Codeberg guys and everything I thought they stood for was genuinely solid.

I did not expect this. I would never have donated if I knew they were going to take a hard political stance against the things that are pushing software development forward right now. This is something else. I am disappointed. I thought I'd have more to say. I thought there would be more nuance here. I really thought they'd have an argument that was better than the internet was giving it credit. It was somehow worse.

I cannot fathom this. There's a really bad horseshoe theory here around the openness of software, wherein the middle, software's open, but as you get more and more private, you get less and less control. And as you get more and more public and free, you end up wrapping around and ending up in the same place. It's kind of crazy that I am more restricted on what I can publish on Codeberg than I am in what I can publish on the Apple App Store.

This is not free software. This is freely available software. You can download it and use it how you please, but Codeberg is not a free software company because they are not honoring the ethos of free software. This type of policing exists outside of any technical or free reasoning. And their attempts to justify this here have a little bit of actual problems that I wish we could have discussed in a more responsible way, just littered with [ __ ] and hate.

It's so dumb and it does a bad service to the people who have genuine arguments against LLMs. Like like some of these people are having their salaries paid by me with a donation that I made for money I got from sponsors that probably had AI on them. I'm going to quote Linus to end this one. AI can be a somewhat painful tool both for maintainer workloads and just from a it keeps finding embarrassing bugs standpoint, but the solution is not to put your head in the sand and sing la la la I can't hear you at the top of your voice like some people seem to do.

The solution is to make sure those LLM tools help maintainers instead of just causing them pain. There's no question on that side. This is so well put and I I knew Linus was based as hell for sending this to the email list, but to have it countered with something this absurd in such a short window really shows how lucky we are to have Linus and Linux. That whole egoism could have went a very different way. And I suddenly now understand way deeper why so few serious people have made the move to Codeberg.

I was genuinely excited to be one of those first, to be one of the people to make the move and advocate for it and get more and more people away from greedy corporations that aren't acting in their users' interests like Microsoft with GitHub and towards a group that I thought cared about their users like Codeberg. But it seems more like Codeberg cares about making political statements than they care about their users because they want to restrict their users' ability to advance software with modern tools.

I'm excited for them to ban Mac users next in order to be more consistent with their rules because the using a Mac is just as damaging as everything they listed here. I think I've said all I have to on this one. I'm genuinely disappointed. I don't know what I recommend now. I guess just host Forge Joe yourself or something. I I don't care. Maybe vibe code an alternative. It can't be that hard. I'm tired. Goodbye.