Most workflows still require explaining context: copy-pasting logs, taking screenshots, describing what’s on screen, repeating yourself and cannot share videos!. The intelligence is there, but the interface is broken.
I’ve been experimenting with a different approach: a native desktop AI (https://wezzly.com) that can see what’s on your screen in real time — terminals, browsers, watches videos — so you don’t have to explain or copy-paste context. The model only acts when asked, everything runs locally, and nothing is stored and its a AI companion that lives on the computer with you with a personality.
Curious what people think: Is continuous visual context the missing piece for AI assistants?