[w10-1]

What does it mean to say Apple is behind in AI or back in 2024, when Apple is not trying to build AI models, and no one else is nearly as evolved as their local architecture?

Is Microsoft's device "vaporware" even comparable to a live demo?

iPhone AI comes via AI-safe pipes: private by design, on device or not, accelerated via MLX. Given that vectoring an AI request to GPU/NPU/CPU/private cloud takes non-trivial engineering, is anyone else as advanced as Apple? It seems wishful to believe interactive UI's will be superseded by untethered, non-interactive (pay by the token) workflows.

iPhone AI is last-mile; the value depends entirely on what you're trying to do. Apple can demo features for their own apps (esp. with personal knowledge), but the real question is what independent developers can do with Apple pipes.

He can say enterprise (not consumers) is the only way to recoup the Nvidia capex, but free local coding models run well enough now on mid-range Apple Macs to make me doubt the business model of Anthropic, OpenAI, and Microsoft, and the upside for Nvidia.

I'm glad Apple will be investing that capex into their own chips.