You Could Run a Language Model on a Bucket of Water (And That Should Bother You)

Sun, 19 Apr 2026 01:04:13 +0000

The math that makes ChatGPT work doesn’t care whether it runs on an H100, a pile of punch cards, or literal ripples in a tub — and sitting with that fact long enough will quietly dissolve most of your assumptions about what “AI” actually is.

I’m not being metaphorical. There are actual demonstrations of mechanical systems — wave interference patterns, physical dominoes, even chemical reaction-diffusion networks — performing operations that are mathematically identical to what happens inside a transformer. Not analogous. Identical. The computation is the same; only the substrate differs.

Compute-Efficiency on BRYSGO

You Could Run a Language Model on a Bucket of Water (And That Should Bother You)