I’m skeptical that local AI will work well with today’s technology. Running capable models consumes too many resources on end-user devices.