@yoozzeek
A while ago, I was prototyping a local ZK rollup but hit a massive hardware wall. Even for a naive state machine, generating a proof locally devoured ~30GB of RAM on an M3 Max and took minutes. The "memory wall" of classical prime-field cryptography makes client-side proving economically unviable.
Why are we forcing CPUs to compute over massive prime fields when silicon physically speaks in bits?
I pivoted to binary tower fields and open-sourced hekate-math, a high-performance, no_std mathematical core for binary tower fields (GF(2^128)). It is designed to bring ZK cryptography back down to bare metal.
https://github.com/oumuamua-labs/hekate-math