RT Rahul | Aerius Labs (@rahulghangas)LFG! S-Two is now accelerated and is performant on Apple Metal.Metal benchmarks are already beating highly optimized simd implementation for trace sizes as low as `log_n=17`, while still remaining competitive with smaller trace sizes.Major metal kernels implemented:– M31/QM31 field ops– Circle FFT / IFFT– FRI fold + decompose– Merkle (BLAKE2s) hashing– Quotient accumulation– Fiat–Shamir channel mix/draw– Constraint VM row eval– MLE folds + circle eval ...