[simonw]

I'm not crazy about this:

> We acknowledge that many developers 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.

Where terms and conditions are concerned I really do value "easy answers".

[what]

Maybe you shouldn’t use a platform that lets the users vote on policy then?

> The second vote was more controversial, but was also accepted with 358 agreements vs 144 disagreements (and 14 abstentions), with a high voter turn-out of around 50% of active members.

[mutkach]

Specific change to ToS: https://codeberg.org/Codeberg/org/commit/96fac426a32d1ba91ff...

> 7. You must not share projects that mostly consist of code written by "generative AI"-tools (including services such as Claude, OpenAI Codex). Such projects having an unclear copyright status (see requirements § 2 (1) 1 and § 2 (1) 3) and furthermore have little safeguards to ensure that they do not include harmful code (c.f. § 2 (1) 5).

[simonw]

The rationale in the ToS change and the rationale in the blog post don't match up.

The ToS change grammar could do with tightening up as well.

[matheusmoreira]

They have no idea what "slop" is. They don't want to have to define it with any precision. Ironically, these rules will be enforced based on vibes.

[what]

They will just put it to user vote, same way all their policy is decided.

> The second vote was more controversial, but was also accepted with 358 agreements vs 144 disagreements (and 14 abstentions), with a high voter turn-out of around 50% of active members.

[matheusmoreira]

No thanks. I get enough of that bullshit in my own country.

[close04]

Maybe they use the definition from the HN guidelines [0]:

> Don't post generated text or AI-edited text. HN is for conversation between humans.

Does this change your view of HN? Of Codeberg? Of vibe enforced rules?

[0] https://news.ycombinator.com/newsguidelines.html

[ceres]

I don’t care about HN the same way I care about my development platform. If I want to show someone my life’s work I don’t send then to HackerNews.

[close04]

You just expect platforms implement “no AI content” rules based on what you care about more. I mean you should leave the platform if you want to create more AI life’s work and they don’t let you, not everything works for everybody, but you can’t come out and object on principle when you accept it elsewhere. This is good for the people who want to signal there’s no AI slop in their work.

And if it’s your life work maybe you want to not hand it to every random platform out there, which can change rules on a whim, and instead host it yourself and own it with your own rules.

[fragmede]

Wait, why am I commenting here then?

[matheusmoreira]

That rule's for HN comments, not submissions. Nobody's getting kicked out for vibecoding a project.

[close04]

You’re trying to make a meaningless distinction to justify holding different parties to different standards: comments are HN’s core content, code is Codeberg’s core content.

HN submissions are links to other platform, so not HN content. Codeberg will let you have links to AI generated content on another platform (like a GH link) so that’s not a worthwhile distinction.

Each platform states a “no AI generated content on our platform” rule or guideline. If you want to resolve the cognitive dissonance that was created when you realized HN does it too, may I suggest using a single standard for both rather than jumping on the weakest way out?

[matheusmoreira]

That's fair.

HN is a social network. Its value is human comments. People come here to see what the tech circles are saying.

A git host's value is in the source code it hosts, not who wrote it. Before all this, Codeberg was all about free software. This has a very precise definition. Free software communities used to gatekeep based on a very simple test: is it free software or proprietary software? The problem is people have started gatekeeping cryptocurrency, LLM usage. The definitions are imprecise, there's no telling where it's going to stop, and it causes needless fragmentation.

[andai]

I was thinking recently about Dijkstra's position on proving the correctness of your programs, and how by his standards, we've all been going by vibes long before AI.

[holgerschurig]

Well, I don't agree that they have NO idea of what slop is. Basically everyone has some idea, see how often many different people here on HN comment "this is pure slop".

Also your claim "they DON'T WANT to have to define" is unproofable. Neither you nor me have any insight into what "they" want or don't want.

I think, for these two though errors your train-of-thought is already invalid.

Further, I'd even argue that precision in such a definition isn't necessary. Especially that the assumed non-existence of a definition would lead to inaction.

But ... if that would imply that we cannot do any action because we don't have a perfect definition, then I again would disagree. In "real life" we e.g. have laws that defines various crimes, e.g. "theft", "murder", "terrorist act" and what not. None of these definitions is perfect. Often they match, but every so often it's not really clear if they apply. In such cases a judge or a jury has to decide.

And that is possible today with things that people perceive as "slop". Someone could just decide. Based on a (imprecise) definition, based on tradition? Like in the judicial system.

There are also many cases where people don't use definitions and still exactly know what's going on. Look e.g. at 5 year old kids. They start to have an idea of property -- and are already very good to know then "this is mine" or "you thieved it" applies and when not. I am however quite sure they can't recite a definition of property or thievedom.

[matheusmoreira]

> I don't agree that they have NO idea of what slop is.

There's no way for anyone reading their rules to know if their projects are slop. It's full of "mostly" and "significantly". That's virtually guaranteed to boil down to "we don't like it".

So "they didn't know how to define it" is about the most charitable interpretation I can muster. Alternative interpretations, such as "the vagueness is a feature since it enables selective enforcement", are also available, and they only get more and more uncharitable.

> Also your claim "they DON'T WANT to have to define" is unproofable.

Had they wanted to define it, they would have done so.

> Neither you nor me have any insight into what "they" want or don't want.

I have insight into their actions, and I'm perfectly capable of inferring context and motives from them.

> I'd even argue that precision in such a definition isn't necessary.

None of this is strictly necessary. Nevertheless, it is desirable that things be precisely defined so that people can know where they stand. Depending on the exact definition, I may or may not be welcome. Therefore I'm not going back there without a precise definition.

[inigyou]

If I showed you an image with black sections and red sections, would you be able to decide whether it was mostly black or mostly red?

[TaLiTr]

I could write a script that could tell you the sum total of red and black pixels.

This is more like asking someone colorblind which shade of red the image is closer to.

[inigyou]

You don't know which parts of your project are AI output? And even if you did know, you couldn't know whether it was mostly AI or mostly not AI?