Gautam Dhameja: running open‑source AI on iOS Neural Engine is ~80% productization, not model work

- X post from Gautam Dhameja argues running open‑source AI on the iOS Neural Engine is production‑ready and that privacy need not be traded off for local inference. - He summarized the engineering reality as '80% work is productizing models,' stressing Core ML optimizations and Neural Engine practicality. - The takeaway is that deployment, UX and integration do most of the heavy lifting for on‑device AI teams. (x.com)