Thanks for playing, and for your curiosity! In honesty I didn't put very good in-game explanation/documentation of what's going on exactly. I wrote a few devlogs, but should've put more info in-game. 

The music you hear is pre-generated by Suno from a soundtrack of ~40 tracks, lots of which I made during/for the game itself (with Suno's help ofc). It just starts at Track 01 for first planet you visit, then Track 02 for the next etc. Not stateful or smart. So you can't "switch harmony" etc. Nor even change tracks in a decent way except by progressing. The music player UI was on the cards but got dropped >.<

When you type in words like "MAXX BASS" etc, those are pinging neurons in real-time on a GPT-2 model I've hosted, that's hooked up to interpretability software that can read out live neuron activations. These are translated back into planetary IDs. So your words = a specific planet. There's a planet for "??" and "a" and "as" and "asdf" and "For what are we if not sailors upon the vast sea?". So you're kinda flying around "latent space" in real time in a way - everything's deterministic, so the universe is finite and navigable (3072 planets in total).

That navigation part I got lost in the weeds a bit on (setting up the interpretability stuff, GPT-2 instance, nginx, cloudflare, oof - not good stuff to crash course a weeks learning in) but I managed to make it enough about finding our dudes a new home that I felt it was just enough to submit. Mostly I just hope people get inspired by the mechanics / use of GenAI in different ways :)