New Category for "AI-Assisted" Software?

[us3r1d]

The one thing I’ve seen @atomozero and @nephele agree on is that the discussion of whether LLM-associated software should have its own section on the forum should be happening in its own thread instead of in all the threads about specific pieces of software.

So.

Can we please add a category for “AI-Assisted” software?

We already have categories for “Native” and “Ports” software, so this seems consistent with existing practice here.

Some folks only want to use native packages, some are fine with code ported from other platforms. Most folks are somewhere in between those extremes.

And some folks are fine with full-on YOLOd vibe-coding whereas other folks don’t want to use anything that’s ever been touched by an LLM. Most folks are somewhere in between those extremes too.

And I think the only way this particular range of perspectives will be able share a forum without regular conflicts is to get the information on whether a piece of software was built with LLM contributions as up-front as possible; giving it its own category does that.

I think “AI-Assisted” is sufficiently neutral as a category title for “software that contains LLM-generated code”.

[Horizons]

I would absolutely welcome a separate “AI-assisted” forum category. (To be honest, I’d be perfectly happy with a “no creating threads about your LLM slop” policy too, but having it in a specific category would be the next best thing. That way the people who tolerate it can do so and the people who would rather avoid it have a way to easily recognize it.)

[Philipingram25]

Honestly, this is 10 out of 10 idea and it could help a lot of beginner coders port a lot of things to haiku that were meant for haiku so we don’t have to rely on Linux slop for port

[zeldakatze]

@us3r1d

Can we please add a category for “AI-Assisted” software?

I’d personally very much dislike it (to say it calmly), as think that it could encourage such software and likely would make the Haiku ecosystem (if you will) feel janky. Personally I’d think that AI-generated software should not be advertised on the forum.

Rushing develoment and creating programs that are quickly ported/generated using LLMs could just create a source of badly working flood of programs which could steer people away from using Haiku at all. It certainly would push me away.

@us3r1d

I think “AI-Assisted” is sufficiently neutral as a category title for “software that contains LLM-generated code”.

“Assisted” makes it sound like there just was minimal input by an LLM. If you generate a majority of your codebase with an LLM, I would not call that assisted but generated by an LLM.

[us3r1d]

@zeldakatze

Personally I’d think that AI-generated software should not be advertised on the forum.

Personally, I agree. But I’m trying to be as neutral as possible in this because there are a lot of pro-LLM folks here too.

@zeldakatze

I would not call that assisted but generated by an LLM.

I’m suggesting that name because it reasonably applies to any software that uses an LLM in its creation, so the category would apply to the widest possible range of software, and it’s reasonably neutral, to keep the forum out of the issue of whether LLM-assisted is a good or bad thing.

To let the individual decide what it means for them and whether that makes it a category to avoid or to give special attention to.

[michel]

@Philipingram25

Linux slop

I see what you did there

[michel]

It’s worth a try.

But the underlying disagreement seems too fundamental. It will just emerge somewhere else.

I’v e made fun of RedoxOS and its insistence that everything must be written in Rust. Well, we now have the Haiku equivalent.

AI is just a symptom. At a deeper level it is a disagreement between those who hold a purist vision of what Haiku is, and those with a more liberal view of what it could be. We saw that during the package management wars (“ZIP files were good enough for BeOS …“), when many of today’s purists were then the liberals. We saw it when QT ports finally got good. And as a yab programmer I’ve long resigned myself to second-class citizen status.

Maybe this crack can be papered over with a forum category. But it will show up elsewhere.

[us3r1d]

@michel

AI is just a symptom. At a deeper level it is a disagreement between those who hold a purist vision of what Haiku is, and those with a more liberal view of what it could be.

I think you’re wrong about this; the objection to generative AI is much more culturally widespread than just among Haiku’s small community.

“Claire Obscure” lost a major gaming award last year over containing generative content; the current Nebula Awards ballot has several asterisked entries on it for potentially containing generative content that will be disqualified if they do.

This is definitely not just about Haiku.

[jackburton]

What happened to Claire Obscure is a perfect example of the insane extremism which surrounds AI usage, since they simply used AI initially to generate placeholder assets, replaced later with assets created by human artists. Oh the crime!

The game is a complete masterpiece, anyway.

[Akakor]

It’s mainly Haiku operating system devs versus users who want a certain app or functionality and now have the opportunity to create and contribute something they need.

[us3r1d]

@jackburton

they simply used AI initially to generate placeholder assets, replaced later with assets created by human artists

That is what they said, though they didn’t replace the assets till after they lost the award from what I’d read?

Yeah, zero-tolerance is obviously an extreme position; you can’t get much more extreme, really. It’s one I agree with and support, personally.

But it’s not my place to push my view of the issue onto the whole Haiku community, so I’m trying to propose a way the community can let each user make their own decision on it without the same fight breaking out every time.

If we have a category for “AI-Assisted” software, and a policy that arguments over the matter aren’t allowed on posts in that category, then maybe we can all share the same forum more peacefully?

[us3r1d]

@jackburton

The game is a complete masterpiece, anyway.

I’ve heard really good things about it, but I haven’t had time to try it. I have so little gaming time that I still haven’t finished “Stray”, and I bought that at release.

[zeldakatze]

@michel

I’v e made fun of RedoxOS and its insistence that everything must be written in Rust

I thought that that’s the entire point of Redox? Not doing that would defeat the basic idea behind that system.

@michel

AI is just a symptom.

The main function of a LLM is to generate text that sounds like natural language. The information the text contains is really just coincidental and mainly revolves around a “it sounds good”, not trying to ensure that the final code is correct. This makes LLMs great for keeping people or scrapers busy, but dangerous as LLM-generated code is hard to review. LLMs are good at creating noise and wasting time of reviewers (or in other contexts, make finding information about something happening in the world almost impossible). They are a danger to any kind of information flow, flooding the zone with slop. LLMs (and image generation models) don’t create knowledge, they destroy it by generating a plethora of articles and “documentation” that is at best half-true and at worst complete bs. “AI” (Generative models, would be the correct term) is not the symptom but a weapon against knowledge and I don’t think we should support that.

@michel

a more liberal view of what it could be

I have a quite hopeful view of Haiku as a well designed desktop system that is well designed and has diligently written software. Of course you could take a shortcut by using a slop generator, but I’d doubt that the end result would be all that well usable. I don’t think quickly “filling up the gaps” with LLM-generated content is a great idea at all. Both Haiku and the application landscape already have made some huge steps since beta2 imo (That was the time when i started using Haiku again after first having had a look at it in 2013 when I actually still was a child).

[Philipingram25]

I feel like adding AI and letting it make more software for haiku, but not system level Integration is a good thing, especially since haiku needs the apps and software and really needs all the help it can get along with a lot of people not really knowing how to code. It’s either that people choose three ways the port or make a software or app on haiku

It’s either they take the short path and just make a .sh launcher or there learn yab but even then most of the time it’s not ever treated fairly and it’s always overlooked the second method that everybody recommends but it so poorly documented on haiku is learning haiku‘s format and learning literally a large steep learning curve C++ especially since haiku’s own documents haven’t been up-to-date and really isn’t a visual or auditory learning system, it is damn near impossible without me needing to go to like three other forums including Reddit and this one to search through a bunch of things due to the inconsistency of the documents or the hardness of haiku‘s code alone

Cause not even knowing like basic Linux or BSD stuff is gonna help you here even some of the coding experience I came with, was mainly useless due to the fact that I did not know how to code on haiku and all the user guide not being updated or really helpful enough to help me code, to the point where I have to use Google AI overview to objectively search all the ways it can teach me how the code or to see if there’s any YouTube tutorials which is very limited also so generally, I believe this is a good thing for haiku

[BlueSky]

@michel

I’v e made fun of RedoxOS and its insistence that everything must be written in Rust. Well, we now have the Haiku equivalent.

No we don’t. The whole (or at least the main) point of Redox is that is an experiment to write an OS in Rust. I’ve got a slight deja vu that I already told you that in some previous discussion

@us3r1d

Can we please add a category for “AI-Assisted” software?

I think the better idea that has already been brought up several times is to label announcement and discussion about apps that have been created with the help of AI as such. Problem solved.

@us3r1d

We already have categories for “Native” and “Ports” software, so this seems consistent with existing practice here.

Software generated by an LLM can also be native by this distinction if it uses the Haiku API. We already have a few apps that fit into this category.

[waddlesplash]

@Philipingram25

and it’s always overlooked the second method that everybody recommends but it so poorly documented on haiku is learning haiku‘s format and learning literally a large steep learning curve C++ especially since haiku’s own documents haven’t been up-to-date

This is what asking questions is for. We have some documentation which is pretty good, though it’s incomplete, though it may be hard to find. So ask simple and straightforward questions about how to do something or where to find something, and chances are we will be able to help.

There are also other languages to write code in besides C++. There are now Python bindings for Haiku, there are bindings to other languages in the depot, there’s Yab, and so on. Plenty of waits to write native software without needing to write C++.

[nipos]

When I first read the title in the topic list,I initially thought it’s a good idea because one can easily mute (=completely hide,if I understand that correctly) whole categories in the Discourse preferences.
Seeing the “daily piece of slop” being advertised on the forum and all the flamewars around it really made me lose a lot of motivation to contribute to open-source at all.
And NO,I’m not saying that the flamewars against LLM slop are a bad thing,quite the contrary.
I find it important that low-quality code built on top of copyright breach,DDoSing of open-source infrastructure and unprecendented waste of natural resources is fought with fire.
Reading the (non-)arguments of the pro-LLM guys that primarily consist of ignorance made me reach levels of anger I’ve not often experienced so far,it was a almost impossibly difficult challenge not to answer to those ignorantly stupid posts in a way that I’d later regret,and that would probably get me banned.

But @zeldakatze has a strong argument that such a separate category could encourage the promotion of more low-quality slop projects and make it look like that sort of projects is especially welcomed.
It seems a good portion of the Haiku core developers agree that they dislike it and don’t want it in Haiku,and it’s primarily the normal users who fall for the marketing slop of LLM companies that they could do everything faster and better by using LLMs,which is obviously not true.
I don’t see a good reason why the Haiku project should even offer free advertising space for slopcoders.

Historically,the Haiku project has always been about taking time to get things right and writing high-quality code.
This resulted in a system with a amazingly stable ABI and API,no dependency hell like on the Unixes,a consistent interface where all parts integrate well with the rest of the system,and amazing performance even on budget hardware.
I’d say almost everyone is here because they like Haiku,because it does something right.
I think that those high quality standards are a big part of what made Haiku what it is today.
Therefore,I’m for completely banning advertising of LLM-“assisted” software and any other form of pro-LLM advertising from the forum.

The idea of quickly auto-generating some half-working application is fundamentally different to the core values of Haiku - stability,quality,taking time and getting things right.
There are more than enough other operating systems and projects that take such code,why not spend your time there instead if you want to use LLMs?
If you use a major operating system like Windows,you’ll quickly notice the masses of absolutely low-quality garbage programs that are bloated as hell,don’t work correctly and can’t even correctly explain the reason why something doesn’t work.
Even worse: Have a look at Android,how many apps don’t achieve to maintain a somewhat consistent interface,have some annoying blinking ad banner and consist of nothing more than a webview hardcoded to a specific URL?
I don’t want Haiku to become like that!

Of course the Haiku project can’t regulate third-party projects someone puts at Github.
I don’t think anyone ever tried that.
The point is advertising of such slopware in the official Haiku forum,which is part of the Haiku project and where the project can decide what content is allowed and what isn’t.

I believe that banning LLM promo from the forum and clearly stating this could even help the Haiku project gain more quality contributors who love what they’re doing.
The number of projects that still care about code quality and the fact that it’s handwritten is decreasing overall.
Haiku could become a excellent example of a project that sticks to high quality requirements and valuing human work in difficult times.

[humdinger]

@BlueSky

I think the better idea that has already been brought up several times is to label announcement and discussion about apps that have been created with the help of AI as such. Problem solved.

The charm of a separate category is that it’d be easy to filter out all its posts. I’m not sure it’s possible currently, they’d still show up in “Latest” and when searching you can select “All” or one specific cateories, but not “all categories, sans x”. Maybe the forum software can be tweaked for that.

[us3r1d]

@nipos

I believe that banning LLM promo from the forum and clearly stating this could even help the Haiku project gain more quality contributors who love what they’re doing.

I personally agree with you that this would be preferable, and would be better for Haiku in the long run.

I just don’t think that this view has enough support in the Haiku forum community to win this outright. And without some formal policy, I think the arguments will just keep cropping up in every thread that mentions generative code.

Would you accept a forum category as a compromise?

[nipos]

Having flamewars in every topic about LLM-generated trash is a reliable way of discouraging people from posting it here,even if it’s not formally disallowed.
So I’m more for leaving things how they currently are if we can’t get it banned.
Otherwise,as explained,it might result in more LLM slop being posted,which is probably not what we want to achieve.
I mean,it’s really annoying to read that,but having the whole forum drown in slop instead,even if I can personally hide it in my preferences,isn’t a good solution either.