Evolution

Dodgy neural-net implementation.

Red dots (bugs) get inputs fed to 20 or so neurons. Neurons take an average of a selection of other neurons, apply a modifyer and return the result. Bugs have two output neurons, which feed 'thrustX' and 'thrustY'. The sim will spawn either random bugs or 'best ever' children, at a 50/50 ratio, until there are at least 20 bugs. Bugs reproduce every 111 'age' Bugs can't die between 0 and 100 'distance' Bugs loop back to 0 distance when they hit 1000. Black dot's appear at a frequency relative to how many bugs are alive.

All behaviour you see below is emergent. Bugs are not coded to jump when near a dot. This is emergent behaviour due to natural selection.

Sim will run at 'realtime' when a live bug gets further than any bug in history, and will jump to hyperspeed when it dies. Hyperspeed runs as many ticks as it can in 50ms per timeout defer. Evolution progress is cpu-bound.