Maybe improper Node.js and/or NPM Haiku port that is not adapted to use Haiku paths.
I don’t want to comment further on the negative tone towards AI criticism in the initial post,and instead add some more facts to the discussion.
I recently found a very long and detailed post that,besides the ecological and copyright arguments,also shows how AI actually reduces productivity due to the many mistakes it makes and how the financial bubble could lead to a huge financial crisis.
Each argument leads to studies and proofs,which makes that an interesting read that allows you to dive deep into the details.
4 Likes
Frankly, as a user of Haikuports, I’m fine with whichever way they choose; as you say, it’s their system, their call. I’d prefer no one used LLMs, for lots of reasons, but it’s not my place to tell anyone that they can’t.
But equally, they really need to stop trying to tell me I can’t _not_ use the software they produce with it. Tolerance has to go both ways.
I’d just like a “made-with-LLM” flag in Haikuports so I can ignore anything with it.
I don’t want to tell any LLM fan that they may not make software that way, but I want nothing to do with it and the most convenient and effective way to do that is a filterable flag on each package.
4 Likes
Oh wow, was it really needed to light such a fire?
Anyway, here is my opinion, hopefully while staying calm and constructive.
I spent a lot of time since my childhood teaching myself how computer works. I like this field because things are fully predictible and deterministic. I can really get “to the bottom of things”, and have definitive answers. The code works, or it doesn’t. Computers are really good at that, and I think I’m reasonably good at reading and writing code. This is very important to me, personally, and is also how the industry operated until now. You have a bug, you collect data, you find the exact cause, you fix it, and it doesn’t come back.
LLMs used to generate code throws this away. At the core of it is a random number generator. The answers are dnfferent each time, and, more importantly, they are not based on truthness, but probabilities. My world view is based on science and facts. These things are not. They don’t care about facts. And with the very central place of the tech industry, I am very worried, also because this isn’t the first place where facts-based systems are being attacked. I am bracing for an obscurantist age. I hope Haiku could be a safe space, a little bit away of such problems. But it may not be.
And that’s just one of the problems. I read here arguments saying that writing software using LLMs is easier and faster. Let’s assume this is true (I can’t begin to understand how this can be, but we all have different brains, so I’ll assume it works for you somehow). Even then, how is this acceptable given the damage that’s being done? Copyright of the training data is being ignored, which should be clearly illegal, and no one cares. Big tech companies have given up their CO2 reduction targets, which will cause massive population displacements due to climate change, and no one cares. M, craft of writing software is being replaced by a partly automated process, but let’s not forget the underpaid people in 3rd world countries, tagging the data for training the models, sometimes manually writing the replies of the supposed “AI”. No one cares. The open web is dying because the training of these models consists of massive amount of requests on every website, making it unsustainable to host a website. I have spent hours tweaking firewall rules and scripts on my hersonal server. Kallisti5 has done a ounch of work keeping the Haiku infra online and making it resilient to these attacks. And then you dare come here and mock us for the lack of progress on Haiku?
I would very much prefer anoutright ban of anything LLM based. But I’m too tired to fight. So all I asked is a label on applications using this tech. Think of it like food labels for allergens, or simply indicating vegetarian/vegan options vs ones that contain animal products in a restaurant or a shop. Is that too much to ask? Would you mock people because they have a plant based diet and as a result they can’t run as fast as you? All while using their work? Can you understand how disrespectful this is?
Pease leave us alone.
15 Likes
Please no left wing propaganda here, really. And data centers are not major CO 2 emissions cause actually.
4 Likes
- Big tech companies abandoning their CO2 reduction emmissions because of AI: https://www.forbes.com/sites/jemmagreen/2024/08/29/why-big-corporations-are-quietly-abandoning-their-climate-commitments/
- AI datacenters CO2 emissions are quickly catching up with airlines emissions: A.I. Data Centers’ Carbon Emission Rivals Commercial Airlines: Study | Observer and more than 2% of emissions in the US. And given the current path for CO2 emissions, every percent we can save is well worth it.
- Climate migrations: Climate refugees: The world’s forgotten displacement crisis | Concern Worldwide
I hope I don’t need to include links for proofs that climate change is linked to CO2 emissions.
Which part of this do you consider left wing propaganda?
8 Likes
I totally agree with everything you say,as always when it comes to this topic.
I’m just wondering why this post is a reply to mine with a link to a article that says the same things plus a dozen more and links to further proofs.
I’ve said these arguments so many times here in the forum already and didn’t want to write the big text wall over and over again,so I instead linked to something that basically says the same and more.
I also had to invest hours of fighting to keep my servers up while they were literally DDoSed by the scraping bots of the AI companies,at their last attempt even with faked browser user agents and coming from residential IPs.
Just to make it absolutely clear: I’m totally against the overhyped fraud that they call AI and prefer to see that banned.
1 Like
I had to write a reply myself from my personal point of view. And also there was already a lot to read in this topic when I checked the forum yesterday, so I didn’t click all the links. But I will read it now 
1 Like
This is my opinion, as a former AI skeptic back in 2023 and now experimenting with recent AI models.
LLMs, if used the right way, can be empowering for small projects trying to grow. With the latest models, I never have to write things like ELF loaders from scratch again - things that I have done so many times - and focus on the real innovation.
For Haiku, you can think of things like adding an automated test framework, improving existing test cases, and other things that do not affect the core system and end-user experience, but are critical to the development process.
This is also what I enjoy when building my projects or building with Haiku - fixing bugs, solving these little problems, finding the exact one line (or even just one character) that caused the issue, then moving on. But what if there’s nothing to fix, and you spend days building something from scratch because your requirements are just slightly different from that of existing solutions? LLMs can bridge that gap - but the final touch is still made by a developer. The art is going nowhere.
If it weren’t for the copyright concerns, I’d advocate for a “human in the loop” policy like what LLVM adopted. This would sufficiently discourage “AI slop” while empowering contributors who use the tools responsibly.
3 Likes
Where exactly do you see “left wing propaganda” here? Why even use these terms in a discussion like that? If you have evidence that contradicts Pulkomandy’s statement, just present it or post a link. Climate change doesn’t care about political opinions or affiliations.
Back on topic (halfway): Here’s an interesting article by MIT Technology Review, that tries to estimate the energy consumptions of current AI (LLM’s for text generation, and diffusion models for image and video generation) and also offers some predictions for the future. The most troubling information I take away from this article is not even the numbers they present, but actually what we don’t know. Because there is apparently next to no public information available about the energy consumption and climate impact of datacenters used for AI. I’m not naive enough to think that will change anytime soon but we should definitely demand that the information should be made public.
3 Likes
“We need to save the planet!” thing is politics and left wing propaganda. Haiku is not about saving the planet, it is off-topic here.
Talking about LLM code contributions to Haiku a bad because it reduce code quality is valid discussion. “Save the planet!” “Save Internet!” is not, it is pure politics.
1 Like
It do not matter. You do not pay that electricity bills. Haiku Inc. do not pay that electricity bills. It is not your or Haiku Inc. responsibility. If Haiku will or will not use LLM agents, it do not change anything.
While I slightly agree with you on the propaganda part, I have to disagree here.
Even though you don’t have to directly pay the bills, increased energy consumption will push up the market price, which will reflect in the bills of other households and businesses - but still a price I’d pay to accelerate development.
What I think really matters is the copyright gray zone - until a strong legal framework is defined, it is a bit legally risky to use LLM-generated code.
Not that OpenBeOS is not in the copyright gray zone already anyway.
2 Likes
I do not have to pay those electricity bills,that’s correct,because luckily we don’t have many AI datacenters in Germany yet.
Those unfortunate to live near those datacenters in the USA (and having lived there decades before the datacenters were built) have to pay the bills.
https://www.bloomberg.com/graphics/2025-ai-data-centers-electricity-prices/
I think that the companies who need the massive amounts of energy should pay for them,but the truth seems to be that datacenter companies get subsidized because growth,and the normal people living there have to pay the price.
1 Like
Good luck with continuing development of your OS on a planet where humans cannot live, or will at least be struggling and fighting for the few stil livable places. I think people will have more important problems than developping Haiku then.
You chose to ignore these problems, and focus on more short term ones.
One of Haiku’s strong points is running on old computers that no other modern OS cares about. This connects to permacomputing, to reducing resources, and, at least for me, obviously to saving the planet.
As for the Internet, well, I wouldn’t be here without the internet, and I believe the project simply wouldn’t exist. This is a space for collaboration that enable such crazy projects and allow them to live for so long. And we should take care about the infrastructure that sustains us. If you ignore that, the very base of everything, projects like Haiku, which are on the outskirts of things, will be the first to collapse. Because the developers will have more important things to do, and there will be no space for this.
I wish for a world where we can all peacefully sit in front of our computers and write open source code for fun, try to make things work right, and collaborate on it from all around the world. Maybe you are in a privileged position where you will not have too many problems continuing to do so. But that is not the case for everyone. Yes, apparently, caring about humans having a decent life is pure politics now. For me it should be basic decency. But now I have to defend it, and you can be sure that I will do my best. If that is not acceptable here because of a “no politics” rule, I will find other places where my views are acceptable.
What about the extra charges for Netlify hosting because of the bots DDoS’ing our infrastructure? What about the raising costs of hosting overall, in part because of the extra traffic from the bots, and in part because of the raise in cost of RAM and other hardware components? We are, very much, paying the bills directly.
This is again a rude attack on our work based on no evidence. Strikethrough typography doesn’t help.
We have spent consiredable effort making a clean room implementation of BeOS. Why pretend otherwise?
Again, facts do matter. Please point to specific problems if you know of any.
7 Likes
Yes. Code quality and law compliance are valid points. CO2 emissions are politics. But for 3rd-party Haiku software it is responsibility of software author, not Haiku team. @anon11892322 can freely choose use LLM agents or not for his work. And users are free to use @anon11892322 work or not.
3 Likes
Even for worst scientific predictions of climate change, humanity will definitely not extinct, not reduce population dramatically and civilization will not disappear. Place where I live will not go under water.
Some people may consider that reducing consumption is worse choose than let global warming progress as is. It is political decision, not pure science. Role of science is to analyze, make models and predictions, but not to make chooses.
It is questionable, that reducing CO2 emission is even possible considering geopolitics and world power balance.
2 Likes
I agree with @X512 here. If we start banning software on HaikuPorts for political reasons we put ourselves on a slippery slope
4 Likes
I feel like this is starting to get very political and very environmentalist which I am for but right now we have to get back on topic with the AI used or not
1 Like
It is reality that is needed to accept and to deal with. For example install Anubis. Internet is quite anarchic thing where various malicious actors existed even before AI bots. And because of various jurisdictions, it can’t be addressed.
Usage of LLM agents for Haiku code will not change anything.
2 Likes