Stop Feeding the Machine: Time to Go Closed Source | Hacker News Reader

Stop Feeding the Machine: Time to Go Closed Source

medium.com

6 pointsemdashslop18 comments

Two trends to watch in coming months/years:

1. Companies will revert to closed-source and cease contributions to F/OSS projects. Their legal team will advise them that trade secrets are the only way to protect LLM-generated code.

2. Contributors to F/OSS projects, and developers who run them, will falsely claim authorship of the code written by LLMs.

These are already the most common reactions to the current legal status of LLM-generated works. If it is "uncopyrightable" and goes into the Public Domain or equivalent, then the F/OSS ideology will be considered untenable, except where software is exclusively written by hand.

This is one major reason that Codeberg, Linux, and other projects have said "no AI contributions or repos", because there is no value in being a Public-Domain-Dumping-Ground. If there is no copyright to assign, sublicense, or transfer, then there is no value in the content to any hosting service.

This is an excellent point. OSS licenses only work because of copyright. The intention was to kind of "judo" copyright law to benefit sharing. If the work is uncopyrightable, no license of any sort applies to it.

Before I started running with the OSS crowd, I wanted to put my code in the public domain. I quickly learned that isn't a thing with US copyright law. Any time you create a new work (of the sort that isn't specifically excluded from copyright) it is immediately and automatically copyrighted by you in that instant. If you don't want it to be, tough. It's copyrighted. The best you could do was apply a permissive license.

But now, perhaps there's a workaround for those who really want to be something they made in the public domain: claim it was machine generated. There's an interesting thought...

SQLite has been public domain for a very long time, and has been very widely used on that basis, with no problem. There are some other countries where this is true.

> Linux […] said "no AI contributions or repos"

On the contrary: https://lore.kernel.org/linux-media/CAHk-=wi4zC+Ze8e+p3tMv8T...

Linus is not GNU/Linux. Nor is he a sole dictator in the project

so who/what countered Linus on this and where is the discussion or final conclusion?

He's implying that other projects in the Linux ecosystem have banned AI. Which is mostly untrue, but I'm sure there's at least one out there.

thanks for clarifying.

> Contributors to F/OSS projects, and developers who run them, will falsely claim authorship of the code written by LLMs

It's only "falsely" if the courts rule that LLM-written code is not copyrightable, and so far there's no indication of that happening, although I do see a lot of people secretly wishing for that to be true.

It is conceivable, albeit unlikely, that courts will eventually find that original authors retain ownership or rights, that is, the ingestion of the corpus by models has not diluted copyright status. Or that the company who publishes the model enjoys some ownership or rights over their output. Or that the artificial nature of every work must be acknowledged or documented in some way.

So to clarify: publishers are already claiming exclusive authorship and exclusive ownership of copyright. The exclusive authorship claims are especially problematic, even today, when major platforms are requiring tags or acknowledgement that there is genAI content in any given work.

If you went on Etsy, or Fiverr, and commissioned artwork, would you be OK with a seller claiming 100% ownership and copyright on a work? What if they assigned, or attempted to assign, that copyright ownership to the buyer?

I'm sorry, you're going to have to dumb it down for me. Are you saying it's "conceivable but unlikely" that courts will find that LLM generated code is owned, collectively, by the millions of people who's code was used in training data? Or is it's "conceivable but unlikely" that LLM generated code will be owned by the person who's prompts elicited the output?

Linux has not said "no AI contributions or repos". In fact Linux has a policy explicitly allowing AI.

I'm not sure the author understands how AI learns. At this point, the fundamental "coding" knowledge is baked in, leaving little to "learn" from more code examples. The real power that coding models have comes from reasoning, testing code, and having broader integration into hardware to develop it. In other words, we've already zipped past the transfer learning phase of learning how to code.

Perhaps, but IMHO that’s too strong a claim.

I agree that better reasoning, testing, and agentic tooling matter, but coding models still benefit from new, high-quality code and real-world coding tasks. Software also keeps changing: new APIs, frameworks, languages, vulnerabilities, and engineering patterns appear constantly. I'm skeptical that “Coding knowledge is already baked in” is supported by the evidence.

What is the evidence that the models are already past the "transfer learning phase of learning how to code" more than they are past the "transfer learning phase of learning how to produce high quality animation or music?". The fact that code is produced is not enough evidence that "machines can't learn more from human work".

In other words: What evidence shows that code has uniquely reached data saturation? As far as I know, existing controlled studies still show gains from additional and better-targeted code training and I'm not aware of studies that demonstrate the opposite.

That's fair. My point is that most of what coding models need to learn to write code proficiently is largely baked in. things like syntax, useful code idioms, data structures and algorithms, important engineering concepts and design patterns etc... are already learned. Code, generally is easier to "bake in" in some ways than creative work because it is highly structured with known and absolute rules, whereas creative work thrives on the surfaces where rules and conventions are violated in some way. I work a lot with Claude Code, not just building, but analyzing how it builds and it is at the point where its failures largely arise from a misunderstanding of my instructions or implied goals, and in some cases a failure to understand basic reality, not on syntax or logic. In other words, it implemented the wrong idea really well (some clean up by humans required assuming any of it is useful). That said, the creative concept needs to come from the human. Case in point. I have a simple python based synthetic data generator. Claude uses it inefficiently because it is a new-ish repo it has not seen so it reads the lengthy README and ~60-70% of the code in the repo to learn it before using it. I had 2 solutions to this. One, create a CLI or MCP interface for models, or two, just create a skill file. I presented only the first idea and it returned only that it was possible if difficult given the complexity of how one creates influences on certain classes in the data. It COULD HAVE just said "It would be a lot easier to make a skill" but it lacked the creative capacity for this leap i guess. I jumped right to the skill idea two paragraphs into the explanation of the core challenges of creating an MCP and did not bother to read on before I suggested a skill. That is the current code barrier, not necessarily understanding syntax etc...

As soon as I learned that genAI companies were inhaling everything publicly available, this was my response. I removed all my code (and websites, blog, etc.) from the public internet and no longer use permissive licenses.

Doing it was a bit painful, a loss, and psychologically difficult. But I couldn't think of any other way to protect myself from further abuse. I still can't. If/when a real solution becomes available, I'll do that instead.

I do still share my code, my websites, etc. I just made everything no longer publicly accessible and am selective about allowing access.

I wrote this article not because I'm sure about the position and not because I want to celebrate the situation. I come from the camp who believed that open source is super beneficial and a positive movement (and to express the counter view is difficult for me).

I'm just puzzled and confused: why is the open source community not furious with this situation as musicians and visual artists are? Why is it considered OK to continue feeding the machine with such a small push back from the open source communities?

> I come from the camp who believed that open source is super beneficial and a positive movement

As do I. I've been very active in OSS for many years, because of the clear (to me) positive benefits it has for everyone. That's why it was hard and painful to retreat from it. I felt (and still feel) that my hand was forced.

> why is the open source community not furious with this situation

I'm not sure, but I do know that the OSS community has changed from what it used to be. Perhaps there's not as much of a unified mission or purpose as with the fine arts communities? Perhaps (and I slightly lean this direction) the OSS community has different ideals now that are more compatible with this? Perhaps it's a sense of resignation?

I don't know. It's an interesting question.

I do know that there are quite a few OSS folk who are, in fact, furious. It's just not an obvious majority of them, and most of the upset ones I've talked with have a sense that there's nothing they can do about it without withdrawing like I have. Their calculation is that it's still a net good to continue even with this problem.

Reply on news.ycombinator.com