* The text is not actually totally coherent; it'll forget which of the characters is which
* the default InstructGPT can be kind of uncreative sometimes, but regular davinci seems more likely to go off the rails; I think I like Instruct better
+stock phrase "Mark had always been fascinated by women. He loved the way they looked, the way they smelled, the way they sounded"
+
+
+
### Xpression Camera
* it's limited in the sense of noticeably not being a magic mirror (that key fraction of a second behind real time, pixelly, it's just doing basic eye-mouth tracking, so expressions are limited in that sense, distortion if you tilt your head too much), but it gets _close enough_ to being a magic mirror to be really exciting: I blink, the woman on the screen blinks; I open my mouth, she opens her mouth. Amazing!!