I was originally going to write this as a response to this blog post, but I realized once I read it more awake that I generally agree with the thrust. That'll (not) teach me to read posts when I first wake up.

For me, there's a lot more going on as well. The only hobbyist programming space I really dwell in is the open source programming space, which isn't exactly a niche community. I'm also not particularly prolific, nor attached to any one particular open source project as a contributor, just an eternal drive-by contributor.

The introduction of LLMs in the open source space has really been affecting how I think about open source in general, and made me largely fearful for this community that I orbit and contribute to.

A big part of the reason I contribute is for personal knowledge acquisition, and the above blog post really captures a good part of what I don't like about the introduction of LLMs: short-cutting the long process of expertise-building. For me, a large part of the point is the process, not whatever "product" comes out at the other end. On a personal level, it's easy, I can just not use LLMs to generate source code.

But moving beyond that aspect, there are things that do concern me about the future of open source that go beyond my personal feelings.

More minor, I wonder how the idea of "personalized software" is going to affect us. More people building, fewer people sharing, likely more strain on library authors and fewer users/builders of open source end products. I don't think there's any incentive for anyone to open source user-facing apps anymore: who cares when you can build your own tailored to your needs?

Likely it's still fine, since having centralized components does still help authors using LLMs, since it's less code for the LLM to generate (gotta mind the token budget!), and at least from a security PoV, it could be safer to trust a well-audited library rather than vibe code a component that reproduces the old security issues of the field.

However, speaking of token budgets, my biggest fear is that open source is going to just turn into a token dumping ground. How many tokens will you tithe this month to the church of open source? Is your value as an open source contributor now measured by the size of your wallet? By the number of tokens you're willing to invest?

At least from what I'm seeing, a lot of open source projects are really embracing LLM driven development, and as this continues, the inevitable conclusion is that the bulk of open source development will just be optimally donating tokens to whatever projects you care about that month. Open source has long favored people with financial security (people that have the luxury of free time, that don't need to care for relatives or children, etc.), but as all LLM developments concentrate wealth and reward those with deep pockets, so to will the use of LLMs in open source. It seems the future of open source will be wealthy Silicon Valley types that deign to donate their monthly token allotment to the commons.

And thinking about this, I'm just left wondering: what is the point anymore? Why bother putting my time in to struggle a bit, help a project that I'm interested in, learn a little more, when someone with a large token spend can jump in and add whatever feature I'm working on or fix whatever bug I wanted to tackle? Maybe this undercuts my prior point about wanting to work on open source primarily for knowledge acquisition, since if that were my only driver, this wouldn't even be a question.

Maybe the solution is just to buy more tokens.