Really really surprised there isn't more discussion about the background inference service that's mentioned in passing here. If you thought Electron/wrapped web apps were a performance problem, I can't imagine the weight of _also_ running a local AI model that's constantly playing Guess My Age.