Nice talk! I'm unsure about the excessive use of analogies. If you don't have a very precise isomorphism it can be unproductive, but the ones you've shown are cool.

About your turnover on the PvsNP problem, did that have to do with realizing that approximations can be quite good for practical purposes? What's the current state of approximations in the field?

I find it interesting because the only application I know where approximations are complete garbage is cryptography, or "adversarial" applications; is that right? In that case P!=NP yields the best of both words: we can build trapdoor functions for adversarial systems but still solve optimization problems well. I find that picture very convincing for some reason (exact solutions are hard, usable approximations are easy).

That license plate is just awesome.

I hope sometime you come back to the light. Stay strong.

see my response here: https://news.ycombinator.com/item?id=10074890

a world where P = NP is a much darker world.

Dude, with all due respect, that's what the church said when the possibility was raised of the Earth not being the centre of the universe.

"We cannae conceive how it can be bearable to be so" !==> "It can not be so"