But LLMs became possible precisely due to centralized server, their economies of scale, and their capacity to train on user data. All of which are against the core tenets of locality, privacy and low cost (gratis) of FOSS.
Sorry guys, but this one is not for you. Unless you go all in on microchips so powerful that they enable local LLMs, but even that seems to be outside of the FOSS competency as it involves hardware, which FOSS doesn't usually touch (with the exception of raspberry pi's and some libre phones)
This is precisely the sort of political demand for stagnation that I'm talking about. Doing your best to make RMS proud isn't going to make this technology go away.
Sorry guys, but this one is not for you. Unless you go all in on microchips so powerful that they enable local LLMs, but even that seems to be outside of the FOSS competency as it involves hardware, which FOSS doesn't usually touch (with the exception of raspberry pi's and some libre phones)