Developers discuss true use cases and practical scenarios for mobile AI on V2EX

On November 5, 2026, a developer posted a thread titled "Exploring the real demand and practical scenarios for on-device AI" on V2EX. The author had not yet updated Apple Intelligence, but after watching various reviews found them uninspiring, and turned instead to a practical question: what do I actually need AI to do for me on my phone, and which on-device AI functions can't be handled by a PC or a regular app?

The poster rules out several categories one by one. Heavy work such as coding or data cleaning will not happen on a phone; he only uses apps to check progress remotely. Translation is already good enough in the system itself, while polishing and text generation feel marginal, and documents still get written on a desktop machine. Image editing and video trimming may be useful, and background removal counts as a small must-have, though it would not be used every day. Where he sees genuine value is in two areas. First, schedule management: spoken into a watch or phone, a request should be analysed automatically for task type (work or life), priority (important or urgent) and time pattern (continuing or recurring), then inserted into a TODO list. He already approximates this with Shortcuts, feeding dictated sentences to GPT for structuring and then adding the result to Dida via its API. Second, smart-home automation: a "coming home" scene in which lights switch on and the air conditioner starts only when he has truly left the office and is heading home with nobody already inside; and temperature, humidity or air purification adjusted according to his fitness and body data. Given the right data interfaces, he argues, a model could be plugged in and run this today. No follow-up discussion or further coverage has appeared so far.