Main document: Cryptarchia v1 and Post-Quantum Security
1.1 In the abstract, "Harvest-Now-Decrpyt-Later" → "Harvest-Now-Decrypt-Later."
1.2 Section 2.4, first sentence: "Tables 1 and 2 maps" should be "Tables 1 and 2 map"
1.3 Section 8.1: "discusses mitigating 'in-band secret distribution'" reads as though in-band secret distribution is a problem to be eliminated. Bowe's point is about hardening these mechanisms. A small rewrite like "discusses how in-band secret distribution mechanisms must be hardened to preserve privacy under quantum adversaries" would be clearer.
2.1 Fr(sl) is undefined. Section 4.4 introduces the epoch nonce formula:
η_B = Poseidon2(tag ∥ η_parent ∥ ρ_LEAD ∥ Fr(sl))
but never defines Fr. This is the embedding of the slot number into the scalar field of BN254. This should be stated explicitly, especially because under STARK migration the target field changes.
3.1 The "on-spend" framing from BZG+26 doesn't map well to Cryptarchia.. BZG+26's on-spend attack targets Bitcoin-style systems where a public key is exposed during a transaction and an adversary must forge a competing transaction before confirmation. In Cryptarchia, Ed25519 is used for the leader_key — a one-time public key binding a PoL to a block proposal. From my perspective ,there is no mempool-style public key exposure.
The relevant attack in Cryptarchia is: can a quantum adversary observe a block proposal, extract the leader's one-time public key, forge a signature on a competing block, and get it accepted before depth-k finality? The window for this is not mempool dwell time but rather the time for k=2160 blocks to accumulate, which at f=1/30 is approximately 18 hours of wall clock time. This is a much more relaxed window than the BZG+26 on-spend scenario.
I'd suggest either introducing an "on-propose" attack class specific to Cryptarchia's leader election model, or adding a paragraph in Section 2.4 explaining how the on-spend class is reinterpreted in Cryptarchia's context and why the effective window is k × slot_duration / f rather than mempool latency.
3.2 Table 1 clock relevance column. The "Fast-clock strengthens on-spend" annotation for leader signatures is inherited from BZG+26's Bitcoin framing. Given point 3.1, this should be qualified — in Cryptarchia the relevant clock constraint is whether a CRQC can forge within the ~18-hour finality window, which even slow-clock architectures could potentially achieve. The fast/slow distinction matters less here than in Bitcoin.
4.1 Grover's speedup claim needs qualification (Section 5.4). The statement "Grover gives a quadratic speedup rather than a complete break" is standard but incomplete for the Merkle commitment context. For collision finding, the best known quantum algorithm (BHT) achieves O(2^{n/3}) rather than the classical O(2^{n/2}). For Poseidon2 over BN254's ~254-bit field:
The document should distinguish which security notion is relevant for each use case. For Merkle root commitments (C_LEAD, ledger roots), collision resistance matters. For epoch nonce derivation, preimage resistance is primary. The Blend PQ document (if you haven't seen it yet) has a nice treatment of this distinction for BLAKE2b and Poseidon2 that could serve as a reference.
4.3 ML-KEM-768 vs. 128-bit PQ target (Section 6.2). The document recommends "X25519 + ML-KEM-768" for hybrid KEX. ML-KEM-768 targets NIST Security Level 3 (~AES-192 equivalent). However, Section 5.4 and the P2 recommendation target "128-bit PQ security via 256-bit hash outputs." If the system-wide PQ target is 128-bit, ML-KEM-512 (Level 1) would be sufficient, or ML-KEM-768 is fine but represents a conservative margin that should be explicitly justified. Either way, the document would benefit from stating a single, consistent system-wide PQ security level target and then mapping each component to it.