Open source is as much a philosophy as the code being available to download. It's great in many ways that the barrier to software has lowered, people can create the things they need even if they're not technical. However, one side effect is that we loose crowd sourced software - everyone's ideas coming together to make something better. Individual software has it's place, but there's also a massive benefit to a community of people with different backgrounds and ideas working to make something the best it can be. Also even with LLMs there's still a burden to maintaining individual software, and even if you pull in other peoples AI generated code there's no guarantee they'll continue to maintain it. With people who are passionate about a project, you can trust in them to carry it on as long as they can - and instead of spending that money on tokens, you can donate to them.
I also agree that LLMs have a place in Open Source - I think this post beautifully extends the message that Daniel Roe posted about (https://roe.dev/blog/using-ai-in-open-source/) - not letting an LLM speak or think for you. I'm perfectly happy to put time into reviewing AI code as long as the people behind that PR are, well, people. Tools like AgentScan are helping reduce the burden of the spam that doesn't follow the contributing guidelines, and giving us time back to review real PRs.