Look, the fact that I consider marking LLM-based software as an (appropriate) stigma doesn’t affect you; you disagree with me. Since you think it’s acceptable to use LLMs and that a user should trust their output, you should be proud to have your stuff marked as LLM-based! It marks you as one of the hyper-productive, up-to-date users of cutting-edge technology!
Or whatever; since I disagree with all that I’m not the right person to list the positives of LLM-use since I don’t think there are any and it’ll just come across as sarcastic.
I saw in the other thread where you said you weren’t up for figuring out how to set up a pro-LLM ports repo; here’s an offer: I’ll help. I’m not a professional developer, but I’ve been a sysadmin since the ‘90s and I built the ReOS ports system for BeOS back when Be went out of business; I’m sure I can figure out how to make a repo on Codeburg work with the Haiku Depot app.
Just tone down the attacks on the folks who don’t want to use LLM-derived code. They’re not just having emotional, irrational responses; there are plenty of legitimate, rational reasons to avoid that technology, and most of them have been mentioned in this thread; the fact that you disagree that those reasons are sufficient is fine, make your own choices for yourself, but the flip side of other people respecting your decision is respecting that they have the same right to make decisions for themselves.
Look, I flagged your post because it is inapropriate to insult people, especially because of their language. Would you have preffered a warning in addition to that? OP has already been offically warned and banned since the original post, so no they are not free to insult anyone either.
Yes, they have been behaving inapropriately, but we don’t need to all stoop that low!
This discussion has been heated enough already.
That’s fair and all, but I am still waiting for the original post to be redacted and for all attacks to the porters to be deleted from there. They add nothing to the “tutorial” and only serve to make people mad.
An apology from the OP about it would be nice as well but if I were to expect that I’d be living in wonderland ngl.
I fear that If i now modify the original post it would not be clear anymore why this was so contentious, and not “just a tutorial”. So I would prefer not to edit it at this time.
This is not just an offer for @Andrea : any of y’all who are pro-LLM are welcome to join that org and I’ll help get a Haiku Depot-compatible directory tree set up in that repo.
When it’s working, I’ll transfer org admin to y’all and withdraw.
Bearing in mind that though the point of the repo is to build LLM-derived code, it would not itself contain that code so unless someone trains an LLM to build haikuports recipe files this wouldn’t matter.
That said, the only potentially problematic bit in the Codeburg ToS is:
You must only share content on Codeberg which you have the explicit right under copyright and other laws to share under the legal terms with which the content is made available on Codeberg.
They’re under German law, and I know almost nothing about that, but my understanding of the current situation of LLM-generated material under U.S. law is that it is not burdened by the copyrights of material ingested by the LLM (though the LLM creators may bear liability) but the output of the LLM is not eligible for any copyright protection.
I AM NOT A LAWYER.
But my layperson’s understanding of the situation would be that this use would not violate their ToS.
No, they definitely didn’t apologize for their contempt toward all the people spending their time porting things on the platform.
The closest thing I can see to an actual apology is this:
And “no” is not an apology. They were asked nicely to remove the ragebait and they didn’t so they clearly don’t give a crap about people trying to contribute to the platform with ports.
I don’t see anything that clarifies or even remotely justifies the insulting tone of his original post towards many contributors’ porting efforts. And no, wanting to have a “provocative tone” doesn’t count.
But I might have missed some post along the way? In that case please point me to that.
It just occurred to me that haikuports builds .hpkg files from the recipes remotely and then distributes those to the Haiku Depot app. I hadn’t really thought about it, and I’m used to servers where the ports tree does local builds. Oops.
I’m willing to help, but not that much.
(That also potentially makes the Codeburg ToS issue more problematic if .hpkg files are being served.)
Whatever; I’ll still hand that Codeburg org over to any pro-LLM devs and help them set up a locally-building ports tree in it or go for full Haiku Depot compatibility if they want to pay for integration and build resources.
Point 1 The work done on porting is beyond dispute.
Yep, and your work on your ports are nice as well.
Point 2 I… However, claiming that USB webcam support is limited to a single model via IP streaming is, frankly, disappointing to hear from you.
Ref: Shhh... Haiku isn't blind anymore! - #42 by Andrea
You actually posted your webcam driver work a few months ago… you mentioned patches to get it fully working. So, is the point self-defeating - even with use of the AI or large model?!? Hmmm…
Point 3 … “branding” or gatekeeping third-party software, …is something I simply cannot accept.
You don’t have to drink the Kool-Aid or eat red meat. Some people will not trust AI, self-drivng vehicles, or robot assistants. They’ll trust children toys from a sweat shop much faster than trust AI. Provide your own solutions to solve your own problems (or for those users welcoming your help). Don’t burn fire in a paper house. Children are still sleeping.
Well, Points 4-7 are answered with the answer to Point 3. You cannot please those people whom do not want the pleasure. So, you can be wiser than the ego and let this go away or fight the Kingdom of Rome till you are blue in the face - or whatever.
I feel like this discussion has gotten out of hand, so would it be better if it was just archive or deleted because this has caused way too many problems and it somehow got the most replies for the wrong reasons
I found this thread one of the hardest ones to read on this forum, because usually things are pretty laid back around here. I learned a lot about the potential legal dangers posed by incorporating AI generated code, and the additional care that needs to be taken with ensuring memory protection in a C++ codebase as in Haiku, etc. So the thread was for me educational, and so I think it should not be deleted.
But it might be best if @Andrea would break out the How to install Claude Code on Haiku bit into a separate post with a simple title such as “[TUTORIAL] How to install Claude Code on Haiku” and change the title of the original thread to something like “[FLAME WAR WARNING] Why Haiku should embrace Vibe Coding to speed up development” to better describe the thread.
It might be good also to amend the original assertion that Haiku has no such-and-such app even after 20 years of development to something like “no Haiku-native such-and-such app…” to take account of all the ports like IceDove etc which make Haiku so useable today. The way that the original argument was phrased certainly was inaccurate, and makes the assertion disrespectful of all the contributors to Haiku, and certainly feels like a straw man argument.
Personally I do not need such labeling (maybe except project is 80%+ generated using LLM?) and investigating for LLM-generated code is extra maintenance burden. If all projects that contain at least one LLM generated commit must be marked, I am against this.
Personally I do not use LLM generated code in my projects.
A couple of weeks ago, Mozilla posted this about their recent collaboration with Anthropic to find and fix vulnerabilities in Firefox.
In this case Claude was able to discover a number of vulnerabilities in Firefox’s JavaScript code which Mozilla’s engineers were able to reproduce and write fixes for. This seems like a good use case for how AI might collaborate with programmers to protect users from malicious actors.
Mozilla has historically led in deploying advanced security techniques to protect Firefox users. In that same spirit, our team has already started integrating AI-assisted analysis into our internal security workflows to find and fix vulnerabilities before attackers do.
