72 comments

[ 0.21 ms ] story [ 22.8 ms ] thread
We are seeing the slow death of social coding. This policy is not inclusive in any way, and paints everyone and everything with the same brush.
We're seeing the death of people writing code. There's no reason for a human to care about what the computer executes, as long as it does the right thing, so why would anyone care about reviewing the code?

Coding is turning into burger flipping -- you ask the computer to do something, then you do a bunch of manual testing. There's a little bit of jargon needed to ask for the right thing, but it's closer to flipping burgers than it is to rocket science.

"Protected by go-away" so I went away.
I don't know if it's not working, I couldn't pass it without making an account (yeah I found out I didn't have a SourceHut account yet)
It took a pretty long time for me (like 15 seconds)
> LLMs are harmful to the free software ecosystem as a whole, as their main use-case in software is to appropriate free software regardless of license and regurgitate it new new projects without any respect for the copyright, provenance, or obligations of the source material.

This is a new one. "Main use" indeed!

Well, yes -- obviously; it takes a huge amount of training data and generates new projects. I guess you could argue that it's not just for new projects -- it's used for old projects too?
New projects aren't just ripoffs of open source projects.

And in the case it is, a lot of open source projects are mostly ripoffs of commercial projects, which are ripoffs of older (now non-existent) commercial projects. It's ripoffs all the way down.

But like.. the idea that it's "most" of the usage is absolutely absurd on its face.

Yikes! I've been a software engineer for over a decade, and I've never had more fun in my career. Running an AI assistant on a cloud machine to pair with me, explore my ideas, learn new tricks, and even write code that I would never write due to timelines and deadlines has been fantastic.

Using coding agents to search for patterns, understanding why our abstractions might not be working because of how users integrate, etc etc.

Not everything vibe coded is bad software, not all bad software is vibe coded.

Why would you ever care about looking at code? Code forges aren't really a useful tool going forward. As agents replace people writing code, the features of code forges really aren't that useful. Nobody will be reviewing agent code other than agents, and you don't need a pull request and code review model for that.

Coding is just going to be asking for a result and manually testing it, no real thought about what code has been generated, and no human should ever need to look at or review it. Even some versioned tarball on s3 will do for the storage.

So, really, who cares about what they do when their entire business model (and Github's) is speedrunning becoming obsolete?

I find it fulfilling to write and read code. It’s also a powerful technology. That doesn’t just change just because there’s another technology more powerful in certain ways. These are some level 0 reasons to want to read code, but there are lots.
Ok, sure. Hobbyist purposes excluded.
You don’t buy the argument the open source movement has been putting out for decades about code transparency increasing freedom?
The code is so cheap today that you can just ask an LLM to generate it for you, for all but the largest projects. I don't get why you'd care about humans being involved, no.
The code is cheap, but the mistakes in production are very expensive!

You want to be leading the agent, not lagging the produced code.

What does vibe coding mean to you? I use the term vibe code to mean software produced by conversation with an agent, with minimal or no understanding of the code / project structure by the person.

Your process doesn’t sound like vibe coding to me, though maybe this part is “and even write code that I would never write”. If you’re just generating snippets as black boxes and putting that into a project you’re otherwise maintaining general understanding of, this isn’t what I refer to as vibe coding. (Not that I’d use this process in a community with clearly stated rules against AI use.)

The people in support of these bans generally aren't interested in "good vibe coded" software. They oppose LLMs altogether on various grounds, not just the output quality.

FWIW, if an LLM is writing significant code, I'd consider it vibe coding regardless of your understanding of the code or expert guidance and almost certainly ban-AI people would too. The scale people seem care about when this comes up is code percentage that's LLM-written to hand-written, not code ignorance to full understanding by the author.

Does the tab completion feature a lot of people were using before ‘vibecoding’ was coined by the AI dude is or isn’t vibecoding in your view?
Yes, relying on auto complete was frowned upon back then.
> The people in support of these bans generally aren't interested in "good vibe coded" software. They oppose LLMs altogether on various grounds, not just the output quality.

I think you're very wrong.

I think mant of the people who "support these bans" do recognize value in LLM's and use them as tools in the kit applied to their workflows but recognize that rampant abuse and misuse of those same tools threatens to ruin (or already has ruined) many community projects that had previously been healthy and thriving.

And so they find themselves insisting upon unambiguous firewalls as an existential necessity for the survival of communities they may have invested years or decades of their lives into.

Who do we figure out who's right?

I think the term vibe coded is unclear. I use agents to do most of my coding work now. Some might call it vibe coded. I use spec driven development of my own design. While the applications are relatively simple, they are reasonably well architected and maintainable.

Some would say I a, vibe coding since I don’t really write the code. I say I’m not because I’m thinking through the features more than vibe coders.

Right!!! LLMs are yet another step up the abstraction ladder. Before LLMs I would write code in C or Perl; I didn't care about the machine code it eventually was translated into---the compilers/interpreters evolved to handle that correctly and transparently. Note that this was not always the case; when I started to use computers, we all aspired to master assembly like the apocryphal Mel ( https://users.cs.utah.edu/~elb/folklore/mel.html ) --- we had to pay attention to machine code and assembly language, because the compilers often didn't generate performant code and were occasionally buggy;

People say that LLMs are not like compilers because they are unreliable. I think with time they will get fixed.

The original definition, as put forth by Karpathy [1], was that you explicitly did not look at the code ever (diffs, errors, etc.). All development, review, etc. is driven entirely through the LLM agent.

But at this point, the term has been diluted down to basically being meaningless.

https://xcancel.com/karpathy/status/1886192184808149383

Are you actually retaining that information though? You're not writing the code, the LLM is. You might not even be reviewing it properly.
It's hard to compare - prior to Opus 4.5 my code was 95% hand-written, with the occassional tab completion.

Now, I would say 80% of my code is written through an LLM, and the last 20% is usually me reviewing adjusting, etc. The part that I find cumbersome is that even when I make manual adjustments, I still need to tell the LLM, otherwise it leads to context drift.

When I compare these 2 times, I think I remember fewer specific details, but I'm getting better at creating more checkpoints.

I saw somewhere that one of the issues with github is all the commits made by the various AI "things".

The other sites may not be able to afford the infrastructure to support vibe coded projects. Maybe that is one of the reasons Sourcehut has this proposal.

Personally I agree with what codeberg did.

The programming community will continue to schism down these ideological lines, its inevitable.

May everyone find their niche but inclusion does not work when barrier to entry is reduced to near zero.

1 - Where does vibe coding end and agent-assisted engineering start? Where's the line?

2 - How do you enforce this?

Probably they go by volume...

You make 10 commits over the weekend once a month for your fun side project, they won't police it (my assumption they can't and won't).

Your slop machines are running day and night, pushing hundreds of commits a day, day and night... You will get on their radar, take a look, then determine "yeah, this person can go somewhere else".

"we're the real hackers with the real skillz! so we're going to create an artificial hierarchy and bureaucratic process to make sure any poser hackers don't get in" <tips fedora>

That's literally all I see out of these types of moves by people/organizations. A will to power and a sort of childlike insecurity that they might be replaced. These folks aren't ok.

> These folks aren't ok

And people who feel entitled to free compute/storage are way worse.

Exactly; being replaced is inevitable, the machines are going to end up more intelligent than us. At some point in the near to moderate future, selling the output of your brain is no longer going to be economically viable. I hope we'll figure out some equitable society before that, but I'm not going to bet on it, so I'll do what I can to be on the side of the winners.
Interstitial does not work on chrome or brave on iOS. Flagging for unreadable.

If it’s because it’s a source code storage service, that’s just not an interesting function. Run your own on a cheap vps with backups to R2.

As more of these smaller forges get similar policies it makes me curious about the statistics behind vibecoded projects.

My anecdata tells me that most of these vibecoded projects are one offs or abandoned. That’s not to say they don’t have value to the people who put them together. What I’m wondering though is: “Do these have enough value to the community to justify the cost of more infrastructure?”

My guess is the answer is no. Then in which case, if you need a place for your pet projects, Forgejo or Gitea should work nicely for you.

> My anecdata tells me that most of these vibecoded projects are one offs or abandoned. That’s not to say they don’t have value to the people who put them together. What I’m wondering though is: “Do these have enough value to the community to justify the cost of more infrastructure?”

Isn't that true of human authored code?

Sometimes, but definitely not to the same volume.
Vibe-coded projects, where the intent is to replace editing source code as the means of change with editing prompts as the means of change, is one of those situations where the intent of how the project incurs change conflicts with how socially-coded open-source projects have traditionally incurred change.

From that perspective, it makes sense to limit "vibe-coded" projects, since they don't fit in that traditional paradigm.

Now, there are big feelings around vibe-coding and a strong aversion to change, so I can imagine this is a rather jarring experience for folks that want to vibe-code and want to have the dopamine hit of sharing that with other folks in a forum that obstensibly is meant for folks to edit code together.

'They are an environmental catastrophe'

This for me the main source of concern. I am having a blast vibe coding. I understand my job is at risk. But I am totally oblivious to the fact that they are accelerating climate change.

If that's the main concern, wouldn't banning them from code sharing sites be harmful? If a person can create a project using AI and share it, then people can use their code without having to reinvent it. If it can't be shared, then anybody who wants that functionality will recreate it from scratch, and if they use AI to do so then you have now burned ~2x the compute=electricity=impact.
Vibecoders like to make all kinds of things that already exist, often don't care (or notice) if the AI rewrites everything instead of using what exists, and LLMs love to do that.
Okay, you can pick between 90% NIH with some sharing or 95% NIH because you tried to stop the sharing. The amount doesn't matter; you're still causing more AI use.
NIH? Anyhow you kind of cause more use of AI also if it seems that everyone is using it, and thus you should adopt it too.

And nothing actually prevents a vibecoder to clone a repository on SourceHut, they just would be less able to add a ton of crap to it and spit it back on SourceHut (they'll just spit it back on GitHub).

This is silly. It would destroy traffic to their site significantly
What is considered vibe coded and slop?

Where is the line?

Drew draws the line of any introduction of AI in any OSS project. (1.) (2.)

Vim, Linux, OpenBSD have AI in them now, even a small tinge of it, so that would be considered slop.

Is it now with some effort with AI assistance that is acceptable or unacceptable or is it still slop?

Or is it more extreme that an OSS project should absolutely not have ANY AI at all to comply with not being slop?

(1.) https://drewdevault.com/blog/AI-in-Linux/

(2.) https://drewdevault.com/blog/Forking-vim/

I like codeberg approach. If you're a small project that doesn't spam ( a commit a day, a few PR a day) you can be as vibe-coded as you want despite their 'guidelines'.
It seems like they're using the term vibe coded as a pejorative to describe all llm use, which is disingenuous. They want to ban LLMs because of political reasons that I mostly agree with, but the term vibe coding invokes being lazy and not even looking at the bad code you generated. But it's just an email subject, so maybe im thinking too much about it.
I am sad to see this. Drew personally told me that I should use my real personal email to contribute ("Services like Firefox Relay are not compatible with the spirit of SourceHut. You should be using your real email address with the mailing lists") -- which I don't do because I don't trust others with it, let alone the general public -- so I was already excluded, but I still had some respect for the project as a whole. I am rethinking that now, and it is a real shame. All the GitHub alternatives are becoming too bigoted to be worth the trouble, because of these anti-LLM absolutists. Can no one see the middle ground? It's like saying that visiting StackOverflow at all is just as bad as copy-pasting solutions from it.
Ironically this “real email” policy (was not aware of this btw) is just as unenforceable as a blanket no-LLM policy. Is “joeschmoe354266@icloud.com” your real email or not, and how do you prove it?

I don’t even use LLMs, but these puritanical stands are a real turn-off. I got into FOSS because I don’t want anyone telling me what to do, not so I can be the one who tells other people what to do.

OK. After reading the post, it doesn't seem so bad. Although I still don't like their ban on analytics on websites. Otherwise I could've de-googled most of my personal life.
Frankly this policy does not go far enough. What about coders whose thoughts were impure because they used ChatGPT to brainstorm or whose brains are ruined because they used Nano Banana to illustrate their blog? I'll hold off from storing my code anywhere until a more righteous service comes online.
When I wanted to get experience with vibe-coding, and build a gen-AI portfolio, among the decisions about what to use it for and how to do it, I very consciously chose the most AI-enthusiastic, popular Git repository service, which is GitHub.

I was a little surprised that some people were outraged that Codeberg (strong ideology/principles) banned generative AI output.

Now that Sourcehut (craft-oriented?) is considering doing similarly, will there be be similar pushback?

It's laughable to me, imagine a company saying that they would only host projects that were written in assembly. I mean, "sure" you do you, but I will not be doing business there.

These kinds of ideological purity tests and the weird bureaucracies and little fiefdoms associated with them are anathema the hacker ethos.

Did you check the number of open PRs and commit/hour of vibecoded projects? If course they are banned from small community-funded hosts, they eat compute and data for free. Honestly, my vote would have been to make vibe-coded pay per commit, but I guess introducing money is taboo for those communities
I actually would have been fine with this move too, “look, first 5 commits per day are free, after that it’s $cost” would have been fine.

But look at the comments and discourse around this. I guarantee you it ain’t about the money.

Look at what happens in codeberg. Small non-commercial, AI driven and even vibe-coded projects are still on the platform. I don't care about what people say, I care about how they act. And to me, codeberg act appropriately.
> I guess introducing money is taboo for those communities

Sourcehut is a subscription service with multiple tiers of access.

I think the different tiers are just for show, but I could imagine that if they felt there was a subset of projects they liked that were putting extra stress on their system they wouldn't be above differentiation at that point.

I mean, I get it; they've got a dying business model, and unless you're paying per commit, you're not a customer, you're a cost center.
If that company only hosted projects written in assembly and you don't write assembly... then you not doing business there is the desired outcome for the both of you, what's the gotcha there?
(comment deleted)
It could be useful to have repositories that are comprised of only "AI"-generated code and repositories that are comprised of only non-"AI"-generated code

If this happens, one could imagine that the former will seek to use the code from the later for "training AI"

The later might charge the former a license fee

This is fantastic. We need more of this. If you want garbage code, GitHub is the place for you.
Hacker News has become a f*cking vibecoding community, the only post that agrees with the proposal has negative votes
I thought sourcehut was general purpose git hosting, not open source mission focused like codeberg?

obv they're going to have content restriction, but there is a spectrum from "actually illegal" to "wonderful chaps whom I love but happen to disagree with philosophically." Slop coding is neither of those, but is closer to the latter than I'd expect a general purpose host to draw a line.

[delayed]
I don’t vibe code, or use AI much when programming, but if I want to experiment with it, I’d like to be able to track my code in the code hosting service I’m paying for, without my repos being subjected to arbitrary scrutiny or ideological purity tests.

It would be different if we were talking about a service I did not pay for: in that case, they can do whatever they like, and I have little to no right to complain.