As I already talk to myself a lot, I'm way closer to this sort of strategy working than I would have thought before you mentioned it. Though I don't enunciate

As I already talk to myself a lot, I'm way closer to this sort of strategy working than I would have thought before you mentioned it. Though I don't enunciate at all when I talk to myself, so whatever was doing the detecting would need to work on subvocalization. It reminds me of a talk someone gave about this ad-hoc verbal shorthand someone invented so that they could use speech-to-text to code. The talk sounded really funny but the process did seem fairly productive.