OK but then you lose the benefit of all future upstream changes and now you have to manage constant rebasing

Engaging the community and intelligently advocating for improvements is a way to contribute to projects as well, especially if you're willing to use a disposable forks to explore the design space, put forth RFCs, PRs, etc.

> OK but then you lose the benefit of all future upstream changes and now you have to manage constant rebasing

I thought they were swimming in enough money to hire someone to do the rebasing. Or dogfood their models to do the same.

Oh the horror of having to do some rebasing in a world where LLMs do all the hard work for you anyway. And at an AI company with unlimited access to compute resources no less.

You really don't want to do that with PG. Keeping your patches rebased will be a huge pain.

Use LLMs.

No, not for that.

For anything, they probably excel at these sort of tasks.

I use https://gist.github.com/nicowilliams/ea2fa2b445c2db50d2ee650... for this.

You could expose this as a tool for an LLM.