@mrwheat.bsky.social Maybe, but the fact is isn't impossible to run an LLM in the same league as GPT-4o on device, especially with older hardware. Makes sense for Apple to have a dual strategy of some tasks locally, some in the cloud.