Open source is a means to an end, if I can generate an ad-hoc solution that fits my use case more than the often bloated standard open source library for it then I'm all for contributing to the dataset that makes that happen. People who are looking at it from an IP standpoint are less about about solving problems and more about putting themselves on a pedestal of accomplishment just for the sake of getting credit. Especially so for corporations.

I’m still convinced that training a model on GPL code makes the model itself a derivative and requires it to be released under GPL terms.

Yes. It's one reason I've lost interest in OSS completely.

Yes, I do, but it's not going to prevent me from doing so.

I believe that the current generation of GenAI (as a market, not necessarily as a tech) is going to crash and burn by 2027. I also believe that open-source will stay and will keep helping people and that, as the world becomes more lawless and free-for-all, we'll need all the help we can find.

Yes, but we can do Open Source in languages irrelevant to corporates for projects irrelevant to corporates.

The rest is closed source.

I’ve moved to more closed source projects for this reason (just for the fun of coding rather than sharing). Though I suspect they still use private github repos in their deals to microsoft

I've been collecting my downvotes calling that out ever since the advent of GitHub.

And I don't think most of those came from idealistic people without any vested interest in AI business.

It is a bit of a race to the bottom for library components: either you open source and it gets snatched by LLM parties or you keep it closed and good luck selling your wares.

On top of that the open source market will increasingly be flooded with (well intended) AI slop built by junior devs.