Post Fiat Research
Blog
Post Fiat research notes, governance experiments, and technical updates.
Orchard/Halo2 Devnet Research Note
Orchard-style shielded accounting, the June 2026 Zcash Orchard remediation, and the corresponding Post Fiat devnet verifier boundary.
Viability of SGLang Replay: Cross-Hardware
An evidence note on SGLang replay across H200 and H100 NVL profiles: exact held-out packet replay, explicit prompt-drift quarantine, MLX capability control, machine receipts, comparison hashes, and shutdown receipts.
LLM Governance Replay: Cheap Public Verification Before Private Committees
A replay experiment on XRPL amendment history: deterministic LLM replay can make governance review cheaper by making first-pass claims public, typed, hash-bound, and contestable—not by replacing validator judgment.