Sovereign-Mohawk-Proto
Thriving Thriving Go GO Python PY
↑ 50 ?
Sovereign-Mohawk-Proto rwilliamspbg-ops/Sovereign-Mohawk-Proto

High-scale Federated Learning (Goal**10M nodes) with formal BFT verification. Features 10ms zk-SNARKs and a 2026 PQC Overhaul: enforcing x25519-mlkem768 hybrid KEX, XMSS TPM-attestation, and epoch-based quantum-resistant ledger migration. Includes Go node agent + Python SDK v2.

https://rwilliamspbg-ops.github.io/Sovereign-Mohawk-Proto/ ↗
HEALTH ? 80 / Healthy
// solo builder// help wanted byzantine-fault-toleranceedge-computingfederated-learningsecurity-auditgolanglibp2ppqcpqcryptoprivacysovereign-ai ?
02-2026 · today · rwilliamspbg-ops
★11 Share View on GitHub ↗

// readme

Sovereign Mohawk Proto

MOHAWK: Mobile Offloading Heterogenous Adaptive Weights for Knowledge

Sovereign Mohawk Proto is a theorem-guided federated-learning runtime with Byzantine-resilience modeling, RDP privacy accounting, TPM-backed attestation, and deployable observability. “Theorem-guided” is not “theorem-proved end to end”: the Lean development is axiom-free and sorry-free, but most claims are proved about a model of the system, and the Go↔Lean correspondence is empirical. The per-claim status is in the traceability matrix — read it before relying on any single claim.

Seeking External Formal Review

We are actively seeking independent review of the formal claims (Lean 4 theorems, refinement bridges, Ed25519 UF-CMA reduction, MiMC/ZK circuit, convergence and straggler models) and the Go runtime security surface — on a zero budget.