Realistically though, there is a good chance I’ll never read every last word of these, and even if I tried, I’d want to follow cross-references as I went, which would mean a lot of flipping around to avoid material read out of order. The thought of giving up after the letter C, and knowing a lot about cognitive science subtopics that start with certain letters, is pretty tragic.

What I’d really like to do is read a random sample of the book, then read a random sample of the pages left over, recursively. This way, if I quit at any point, I come away with a more or less unbiased sample, whereas if I stick with it, I can be certain I read the whole thing.

So last night I wrote this Python script to help me do that. I haven’t written much in Python, so it’s probably unidiomatic, but I think the implementation is pretty clean. You could probably implement this in about half the number of lines of code, albeit with hideous inefficiency, if you used Python’s set classes. Oh well.

This is a preliminary version of an idea I had about a year and a half ago, which I wrote earlier this year and then forgot about. It’s an automated match cut between an orthogonal projection of a cube and two-dimensional polygons with simulated mass and solidity (thanks to box2d). You can see (occasionally kludgy) code and play with an interactive version here. I’ll be reimplementing this with support for perspective projection in C++/OpenFrameWorks soon.