Remix.run Logo
▲ ricardobeat 5 hours ago

A note to the developers in the project, if anyone is reading: your repositories are in violation of the new Codeberg policy that forbids AI-generated code [1].

[1] https://blog.codeberg.org/protecting-our-floss-commons-from-...

▲lucideer 4 hours ago | parent | next [-]

It's self-hosted on https://code.overheid.nl/ & they mirror a subset of it on Codeberg.

That might still be in violation but I would imagine it's low on Codeberg's radar since they're not making primary contributions there (& if I were DAWO I'd also be interested in limiting direct community agentic contrib).

▲karel-3d 3 hours ago | parent [-]

They have "migrated to codeberg" which is written in the repo itself.

However Codeberg actual "fuzzy" policy is less strict than "no LLM used ever".

▲KomoD 5 hours ago | parent | prev | next [-]

And Codeberg also blocks real people from accessing it, for who knows what reason. I just get "Access denied." with no more info.

But they do let me visit blog.codeberg.org

▲danadam an hour ago | parent | next [-]

They block some user-agents, maybe that's it?

From [1]

> We must regularly implement temporary security measures to ensure the availability of Codeberg's resources and prevent abuse.

> The 403 you have received is expected based on the UA.

From [2]

> What is the version of your chromium? AFAICT <=137 gets 403 Forbidden: [...] while >= 138 works ok: [...]

[1] https://codeberg.org/Codeberg/Community/issues/3084

[2] https://codeberg.org/Codeberg/Community/issues/3078

▲sevenzero 5 hours ago | parent | prev [-]

Had that issue with HTTP accessed repo, switching to SSH solved it for me.

▲mitxela 2 hours ago | parent | prev | next [-]

Codeberg does not forbid AI-generated code. It forbids AI-generated projects. You would know this if you read the policy.

▲llimllib an hour ago | parent [-]

The specific language is:

> projects that mostly consist of code written by "generative AI"-tools

And "mostly" here seems egregiously undefined to me

https://codeberg.org/Codeberg/org/commit/96fac426a32d1ba91ff...

▲WarmWash 41 minutes ago | parent | next [-]

The hardest part of policy isn't coming up with smart policy, it's figuring out where and how to draw the lines.

▲tcfhgj an hour ago | parent | prev [-]

implies for me >50% - instead of >0% as suggested by @ricardobeat

▲gagik_co 5 hours ago | parent | prev | next [-]

It’d make the most sense for the Dutch government (or DAWO) to have their own Forgejo instance if the goal is data sovereignty.

edit: looks like this was originally the case and then they moved?

▲cornedor 4 hours ago | parent | next [-]

They use both. https://code.overheid.nl/MinBZK?q=dawo&sort=alphabetically as the official source, and https://codeberg.org/DAWO for the community parts.

▲mitxela 2 hours ago | parent | prev | next [-]

We can insist everyone self-hosts everything, but is it worth it? Do you have your own email? It's quite normal for a government agency to contract a third party to supply e.g. furniture, instead of chopping its own wood.

▲gmueckl an hour ago | parent [-]

But furniture is a one time sale, not an ongoing service. That's a bad comparison.

▲mitxela 36 minutes ago | parent [-]

A third party supplies their electricity too.

▲utopiah 5 hours ago | parent | prev [-]

clearly shows the need for federated forges, where you get to decide on your own rules but you still benefit from interacting with a broader set of collaborators.

▲amiga386 4 hours ago | parent | next [-]

That's likely to end up Balkanized like the fediverse, where every instance has a list of other instances they refuse to federate with, often because politics or petty drama. Though to be fair to the fediverse, sometimes it's because those other instances are full of child porn and most people just don't want to see that.

It amuses me that Mastodon Central uses corpo-friendly labels for the sites they won't federate with, e.g. (https://mastodon.social/about)

    tsundere.love   Suspended (Hate speech)
    you**.**ve      Suspended (Inappropriate content)
Whereas the more free-wheeling sites, like the ones Mastodon Central blocks, tell it exactly how it is: (https://tsundere.love/about)

    nadeko.pw      CP
    mk.fedcast.ch  Spam
    youjo.love     hosts at least one admitted pedo
▲mitxela 2 hours ago | parent | next [-]

If you want to follow some CP poster on nadeko.pw then you can ask your admin on tsundere.love to unblock it and he will probably just ban you for admitting you like CP.

Where did we get this idea that federated instances should never block anything? Even warez sites have rules.

▲fc417fc802 3 hours ago | parent | prev | next [-]

A lot of such issues could likely be resolved by a better thought out protocol. There are certainly plenty of self righteous people who would refuse to federate purely on principle but more typically the issue is simply that one does not wish to be exposed to or traffic in any way the content in question. A well thought out self categorization system could largely address that need - you'd only need to refuse to peer with those who failed to apply it. AI also makes any such effort easier than ever.

▲numpad0 an hour ago | parent [-]

Doesn't work. People from different geographical areas just can't agree on categories. So even correctly applied labels are useless.

You'd think "no lolis" would be reasonable. It falls apart when people look 14 into 50s, to you and your local police, but not to themselves.

▲Akronymus 3 hours ago | parent | prev [-]

weirdly enough that reminds me of wattpad and ao3, where wattpad uses much nicer sounding terms while ao3 uses as accurate terms as possible. (which makes ao3 seem more "degenerate")

▲poulpy123 4 hours ago | parent | prev | next [-]

That's a recipe for drama as shown with mastodon when they got publicity few years ago

▲embedding-shape 4 hours ago | parent | prev [-]

Yeah, waiting for this too, feels like the final piece of the puzzle. Tracking issue from Codeberg + more general information about the effort around the the ecosystem:

- https://codeberg.org/forgejo/forgejo/issues/59 ("[FEAT] implement federation")

- https://forgefed.org/

▲xvilka 8 minutes ago | parent | next [-]

Given how slowly it progresses in Forgejo, there are better alternatives - Radicle[1] (full decentralization, not a federation) and Tangled[2] (federation).

[1] https://radicle.dev/

[2] https://tangled.org/

▲athrowaway3z 4 hours ago | parent | prev [-]

Final piece of what puzzle?

We can do big federated/distributed development with git at the project level. See the linux kernel.

If the goal is fancy UI then there are alternatives.

If the goal is to do a big multi-git with cross repo issue tracker, then that implies either:

- you're a public site for random repos like codeberg

- you're a special purpose forge and identity management is a big issue you want full control over.

IMO - For this to work, they should go with the latter and not try to offload identity into some federated scheme.

▲NoahZuniga 5 hours ago | parent | prev | next [-]

Their instance is self hosted right, so this doesn't apply I think.

▲ransom1538 5 minutes ago | parent | prev | next [-]

hahaha

▲graemep 5 hours ago | parent | prev | next [-]

Is it on Codeberg? I looked self-hosted to me.

▲KomoD 5 hours ago | parent [-]

https://code.overheid.nl/MinBZK/DAWO-NixOS says:

> DAWO Core has moved from https://code.overheid.nl to Codeberg to facilitate community collaboration.

▲galumptuous 5 hours ago | parent [-]

Does this mean European governments have or will have an alternative(s) to GitHub as well?

▲lucideer 4 hours ago | parent [-]

Forgejo (the Codeberg-maintained Gitea fork) is essentially becoming an EU-based community of familiar compatible alternatives to GitHub.

It's not quite a federated/decentralised platform in the traditional sense, but the decentralised nature of Git makes it a close enough equivalent, & the fact they're built upon a shared platform makes them very easy to navigate. So setting up public instances for each government or org makes a lot of sense.

Codeberg even maintains a list of alternative pubic Forgejo instances[0], some of which are explicitly pro-AI.

[0] https://codeberg.org/forgejo-contrib/delightful-forgejo#publ...

▲rvz 4 hours ago | parent | prev | next [-]

I predict that Codeberg will soon realize that they have no choice but to loosen and allow AI generated code.

There is no point in being on the extreme end around banning all code generated by AI.

Regardless, there is AI generated code in many repositories on Codeberg anyway.

▲mcv 8 minutes ago | parent | next [-]

I don't think they're banning all AI generated code, just projects that are primarily AI-generated. They'll probably also object to PRs that are entirely AI generated. Or at least they will when they become a problem.

▲amrit3128 an hour ago | parent | prev | next [-]

This is less about the code and more about load. It's community funded so allowing agents to willy nilly push and pull will be very expensive.

▲lucideer 4 hours ago | parent | prev | next [-]

I've seen so much commentary on the apparent strictness & extreme nature of Codeberg's approach & no actual examples of it.

They have a strict policy (which many have rightly pointed out is impossible to fully enforce), & their "enforcement" has mostly relied on good faith contributions to date. They're also accounting for the fact the ToS change came without advance notice for many of their users who were already making active use of agents, hence there remains many AI repos (I have yet to hear of any removals).

To me that seems like the best approach one can hope for - a lax policy is a slippery slope (presuming dissuading AI contrib is your goal), & wouldn't satisfy their legal concerns (hosting code of questionable copyright status), & strict enforcement isn't viable in any case (especially at their scale), so this seems to me to be the best compromise.

▲teekert 4 hours ago | parent | prev | next [-]

It may be more a measure against explosive growth than being anti-AI in principle.

▲tcfhgj 4 hours ago | parent | prev | next [-]

AI generated code is allowed on Codeberg, but not if it's mostly ai generated

▲TiredOfLife 2 hours ago | parent | prev | next [-]

People who act out of ideology very very rarely change their opinions.

▲Muromec an hour ago | parent [-]

They can however lose the means to enforce their opinions on others to the point those opinions don't matter.

▲mitxela 2 hours ago | parent | prev [-]

Did you read the actual policy? It isn't extreme in the slightest.

▲tapoxi an hour ago | parent [-]

It's stricter than Debian's, which is one of the most conservative Linux distros out there.

▲docheinestages 4 hours ago | parent | prev | next [-]

Codeberg is trying to solve the right problem with the wrong solution. But hey, it's democratic!

▲lucideer 4 hours ago | parent [-]

I like Codeberg's solution (hard written policy, light good-faith based enforcement), what would you suggest as a better solution?

▲Muromec an hour ago | parent | next [-]

Hard policy, light enforcement in practice means there is an actual shadow policy that you don't know, which is pretty bad to the point that it would be illegal if the government did it.

▲lucideer 10 minutes ago | parent | next [-]

Curious about your comment on two counts (genuinely, not being argumentative):

1. how is this bad, in Codeberg's case, practically speaking?

2. in what cases is this illegal for governments to do? In practice no legislation is perfectly enforced, & even judiciaries have varying degrees of exactness across countries - to the point that things such as desuetude are densely studied legal topics.

(note: I'm aware of the negative ramifications if desuetude but not aware of any negative ramifications of Codeberg's specific ai policy)

▲mcv 5 minutes ago | parent | prev [-]

Governments have a lot of laws that they don't enforce, or enforce only lightly. For example, weed is still illegal in Netherland, despite many decades of tolerating its use.

▲rhplus 3 hours ago | parent | prev [-]

Their contention seems mainly focused on the heavy resource use of AI tooling.

They run a service: they could impose limits on the number of submissions, builds, branches, scans, etc, and give additional usage credits to the large community projects they want to encourage.

▲lucideer 2 hours ago | parent | next [-]

That & the legal aspects of hosting code of questionable copyright.

The LLM revolution has proven that if you have enough institutional investors, the Berne Convention doesn't apply to you, but it's sofar unclear whether that exemption from international law extends to small EU-based non-profits.

▲mitxela an hour ago | parent | prev [-]

But they don't want to. They want to provide all needs for approved projects. They want that if a FOSS project legitimately releases 50 times a day, they want to make it work. If a slop project is wasting those resources that could be used for KDE, on nonsense, then they don't.

They're not a free for all, they do KYCP (know your customer's projects).

▲streetfighter64 5 hours ago | parent | prev | next [-]

Fair note, but I'd avoid the cheap shot "more open". You'll just bring up the old flame wars about whether the MIT-style or GNU-style licenses are more open.

The same applies here, is it really "more open" to allow corporations to profit by removing the openness of your work?

▲ricardobeat 4 hours ago | parent [-]

Point taken, trimmed that. I do take issue with their position but not the place to start a new discussion.

▲embedding-shape 4 hours ago | parent | prev | next [-]

Did you actually read through that post you linked? You realize not every project that was ever touched by a LLM is being thrown off? Read the "Some early, but informal guidelines" particularly, as you've seemingly missed that.

They're targeting slop and essentially spam, not everything that could possibly have been built by LLMs.

▲ricardobeat 4 hours ago | parent | next [-]

Quoting from the post:

    > If you see yourself on this list, you don't need to move right away, but
    there might be other places that better fit your needs:

    - Projects written and maintained with heavy use of LLMs
All of the commits I browsed here seem to be LLM-assisted, so these projects are in this category. The policy being fuzzy is in itself a problem - "likely to be tolerated" is not a great policy to have your entire government depend on.

Personally, the refusal on "Projects heavily tied to the LLM ecosystem, e.g. LLM-written tools to ease LLM usage" is enough reason, as that's a good chunk of what I've been working on.

▲rhamzeh 4 hours ago | parent | next [-]

> "enough reason for me to move out"

And that's not a bad thing. That's why they are communicating, so if you feel that is you and you're uncomfortable about it, you can move out and it's a win win.

Not everything should service all of us, especially not the smaller services. It's OK for places to have their own rules and them being upfront about this is a *good* thing.

▲embedding-shape 4 hours ago | parent | prev [-]

> The policy being fuzzy is in itself a problem

No, the policy being fuzzy is a feature, because they're not 100% against all LLM usage, they're against slop and AI-spam, which is what they're trying to combat.

Laws are fuzzy, and we're doing mostly fine with this, of course there are edge-cases. But that's exactly why laws are fuzzy in the first place.

If you have a "automated software factory" or whatever the vibe-coders are up to today, then yes, Codeberg is probably not for you. But if you sit with a agent TUI in one tab, review things closely and don't just "throw code over the wall" without reviewing it, you have nothing to worry about with hosting code on Codeberg.

▲wafflemaker 4 hours ago | parent | prev [-]

Not OP, but I stand corrected. Thought it was 100% no generated code. Now I know better, thanks! Looks like codeberg aren't as mad as I thought.

▲mitxela an hour ago | parent [-]

In the future, consider reading a policy before complaining about it.

▲schnebbau 5 hours ago | parent | prev [-]

Codeberg are going to have issues in the future when all code is AI generated (except for artisanal code of course, written for the love of the craft but not to achieve anything of merit).

▲Thanemate 4 hours ago | parent | next [-]

>but not to achieve anything of merit

Educating oneself to provide better quality of work still requires writing code by hand, and by extend is an achievement of merit.

▲max_streese 4 hours ago | parent | prev | next [-]

Assuming writing code by hand is slower (which I agree with), why can't it still have merit compared to code written via llm harness?

▲schnebbau 4 hours ago | parent [-]

Because no one is going to hire developers who write code by hand, apart from a few niches. For the same reason that no one hires a courier who delivers by horse anymore, or a miller who uses windmills to grind their grain, or a cooper who continues to use wooden staves and iron hoops instead of forming plastic or metal containers, or...

▲fc417fc802 3 hours ago | parent | next [-]

> or a miller who uses windmills to grind their grain,

Huh. For all the "traditional", "organic", "hand crafted", and other "things" I've encountered that isn't one of them. I wonder why no one is advertising fair trade hydropower stone ground organic flour?

▲tmtvl 3 hours ago | parent | next [-]

Because people generally don't like having bits of millstone in their flour.

▲EraYaN 2 hours ago | parent | prev [-]

At least where I am (the Netherlands) there are plenty of mills still milling.

▲gspr an hour ago | parent | prev | next [-]

Your analogy is terrible.

The act of verifying that parcel is delivered, that the grain is correctly ground, or that the container is up to spec is far easier than the act of delivering/grinding/making in the first place (and in the latter case, there's perhaps even an argument to be had that the modernized process eases verification).

For a lot of coding, this is simply not the case. And so people will retort that that's fine, LLMs will help with verification and maintenance too. I think you're seriously misguided – you're accidentally trying to turn human prose into a programming language, and it's a terrible terrible idea!

At any rate, I find it extremely worrying that people like you keep reiterating this point as if it's a fact. If it's true, then surely we'll very soon see the evidence. The way you speak of the "near future" with certainty, as if everyone agrees, gives me the same sour taste as cryptobros or MLM conmen. Yuck.

▲applfanboysbgon 4 hours ago | parent | prev [-]

Yeah yeah, except LLM code is absolute dogshit and is not on a trajectory to ever not be dogshit.

▲schnebbau 4 hours ago | parent | next [-]

https://knowyourmeme.com/memes/akakichi-no-eleven-redraws

▲dwb 3 hours ago | parent | prev [-]

I’m sorry, but this isn’t the case. I don’t like it, on balance, but managed responsibly and steered well, plenty of LLM code is fine.

▲applfanboysbgon 4 hours ago | parent | prev [-]

Policies are enacted for current reality, not fantasy futures. If your magical fantasy becomes reality (it won't), they can just... change their policy. Of course, you aren't thinking about practical effects of policies on platforms, only on how it's an insult to your LLM-bro way of life.

▲schnebbau 4 hours ago | parent [-]

> If your magical fantasy becomes reality (it won't)

Keep huffing that copium my friend. I've set an alarm to follow up here in a couple of years.

▲gspr an hour ago | parent | next [-]

You sound exactly like the cryptobros of yore. Just stop.

If you're right, the evidence will bear that out. Stop evangelizing. It's insufferable. And it's paradoxical that so many who believe this technology to be absolutely world-changing also feel that the technology needs to be pushed onto people actively. It reminds me of religious people ("my god is all-powerful, but if you meaningless mortal insult him he'll be mightily upset").

▲gmueckl an hour ago | parent [-]

I also hate preachy sermons about new tech with a deep passion and that transfers somewhat to the tech they preach about.

Despite this, there is a reality that LLMs for coding can create pretty decent results fast. There is something real and usable hidden in all that mess. It's like the dot com bubble in some ways: the bubble died with all the overhyped "x on the web" BS that had no real substance, but the web itself stayed. This time around, LLMs and agents will stick around as tools, but the hype will likely die a loud, messy death.

▲gspr 18 minutes ago | parent [-]

Sure. But HN these days is the epitome of "x but on the web" in the AI era. It's insufferable!

▲applfanboysbgon 4 hours ago | parent | prev [-]

Sure. You can come gloat about how you're unemployed because nobody needs a prompt monkey when they can just prompt the LLM to produce dogshit code themselves.

▲schnebbau 3 hours ago | parent [-]

I climbed out of the code trenches years ago - I have multiple small companies, I'll be fine.

And yes, they both have moats that will take AI at least 5 years to erode, at which point I'll just retire.

Regarding the dog shit code comment: this is the problem with your whole line of thinking. Customers don't care if the code is dog shit. They care if they get value from the product.

As I mentioned, there will still be some demand for devs who code by hand, similar to how theres still demand for blacksmiths who make ornamental weapons.

▲applfanboysbgon 3 hours ago | parent [-]

Dogshit code doesn't give customers value from the product. You can get away with enshittifying your code/product if you already have an iron grip on the market and network effects make it impossible to escape, eg. Windows/appleOS/Android, Youtube, Facebook, Twitter, Discord, etc. At that point, you're relying on the inertia of community lock-in to replace product quality. However, this strategy does not work if you aren't a FAANG-scale monopolist. If your code and product is dogshit, you will never get off the ground. You need not only a better product but an overwhelmingly superior product to the competition to defeat inertia. And we see this playing out in reality: there are no vibecoded software disrupting the market. At most vibecoded software is carving out a new MLM-style pyramid of vibecoding tools, but it is not replacing hand-crafted software.

▲schnebbau 3 hours ago | parent [-]

This is just wishful thinking.

Building the product was never the hard part. Getting it in front of people is. That will continue to be true, except the build stage will be even easier.

▲andersonpico 2 hours ago | parent [-]

> Building the product was never the hard part

this is an absolutely insane take

▲gmueckl 36 minutes ago | parent [-]

Is it? A lot of software is really just a pile of context-specific database front-ends. There is virtually nothing difficult in the implementation itself. Most businesses in this space survive because they know how to dress up their pile of CRUD for their customers. And it gets worse: a less reliable, buggier product may still be the better fit for a customer if it works well enough for what they use it the most.