Replay-grade governed-loop primitive for AI agents — pure-TS kernel, zero runtime deps.
Pulled from szl-holdings/platform · packages/codex-kernel (release v1.0.0 → v1.2.0).
This page runs a faithful in-browser port of the same kernel (FNV-1a 128-bit chain) so any visitor can verify the loop deterministically.
Paste a governed-ops payload, or load the canonical Dresden-Venus emulator payload. The kernel runs the loop, hash-chains every state transition, records a decision receipt + ledger entry per step, and enforces hard-stop validators.
— run the loop to see hash-chained state transitions and hard-stop validator output —
— proof_ledger.jsonl appears here —
Recompute the chain from initial_state + trace.jsonl and assert every transition replays bit-identical to the recorded final hash.
— run the loop first, then replay —