[manyatoms]

In 10 years it'll probably be affordable to have 1TB of (v)ram in a home computer.

Will the model sizes keep getting bigger such that the large providers have much of an infra moat compared to local home inference?

[bakugo]

Have you been paying attention to hardware prices recently...?

[giancarlostoro]

If eventually the home models are good enough that you dont even need a cloud AI provider, what happens to their bottom line? Local AI can do local tool calling just like Claude Code can do.