Playing with a strange pen. The ink stretches backwards into the past and forwards into possible futures. The two sides make a strange loop: the future ink influences how I draw, altering the future's past ink…
(it's very interesting how the "feel" of this pen with realtime predictions is totally different from this pen with 200ms predictions, which is totally different from this pen with 2500ms predictions… it's so fertile making discrete systems continuous!)
Optimizing is a great way to map a new system's territory! Current trying to make a non-interactive machine learning systems interactive: a treasure hunt in a strange land!
2.5s -> 250ms: reimplement on GPU
250ms -> 50ms: fewer stages
50ms -> 25ms: fewer textures
25ms -> 2.5ms: ?