I've been thinking about self-hosting (more) AI models, but I am increasingly of the opinion that the opposite of the _You own nothing and you'll be happy._ mantra is not _You will own everything and ...
I subscribe to the idea that the not too distant future there will just be a compute "fabric" of every device mobile to rooms of server racks will be running the model, or whatever we call it after that. Projects like https://github.com/Mesh-LLM/mesh-llm are a glimpse. All the pieces are maturing in own silos until put together well.
11❤️1👀1💜1💯1🤔1