Re: [spam][julia][wrong] Probabilistic Program Synthesis?

thinking on how to quickly expand a sketch-based synthesis system (_that can call itself recursively_), to make an agi with human supervision: basically you want to select the part of the problem that is most helpful for the user to help with, and present information to them in a helpful way - ideally while simultaneously trying things on one's own - but alternatively give them an option to say the information is not helpful, and pursue things further, selecting user presentations that produce reduced delay, that the user won't likely reject or will say they could try if really needed [further thoughts misplaced]