This document is the overview and position document of the post-quantum work. Five documents sit beneath it: a safety analysis and a candidate evaluation for the Blend Protocol, the same pair for the Bedrock layer, and a shared benchmark document measuring the candidate primitives on real hardware. Each of those answers a narrow question thoroughly. None of them answers the question a reader arrives with first: what is the overall shape of this migration, what has to happen in what order, and who decides what.
That is this document’s job. It combines the Blend and Bedrock pictures into one, states the phasing and sequencing that the two safety analyses defer, and maps every open decision to the document or RFC where it will be made. It deliberately carries less detail than the documents beneath it — per-component mechanics, parameter tables, and measured timings stay where they are, and are referenced rather than repeated.
Two things remain out of scope here. Candidate selection is not made in this document: the candidate documents present the trade-offs, the benchmark document supplies the measured costs, and the choice belongs to dedicated decision documents and RFCs. Post-CRQC bootstrap — how a node joining the network after a CRQC exists verifies historical blocks whose proof soundness has lapsed — is identified in the Bedrock analysis and deferred to its own document; it is a consequence of the migration rather than a part of it.
For a reader new to the programme, the intended order is: this document first, then the safety analysis for whichever layer concerns them, then the corresponding candidate document, with the benchmark document consulted for measured costs throughout. The full document map at the end lists all six.
The Logos stack instantiates its cryptography almost entirely with elliptic-curve constructions: Ed25519 for signatures, X25519 for key agreement, and Groth16 over BN254 for zero-knowledge proofs. All three fall to Shor’s algorithm. The hash-based components — BLAKE2b and Poseidon2, and everything derived from them — do not.
The important framing, established in the Blend analysis and carried through the whole programme, is that this is a change at the implementation level, not the protocol level. The protocols need the same abstract primitives they needed before: a key encapsulation mechanism, a signature scheme, a zero-knowledge proof system. Only the concrete instantiations change. The migration is therefore a primitive-replacement programme, not a protocol redesign — which is what makes it tractable to plan in phases at all.
Four properties shape everything that follows:

Figure 1: The two threat classes and the deadlines they impose.
The migration programme is organized around a single distinction: the point in time at which an adversary is able to act against a given primitive. This timing determines the deadline that applies to each surface and partitions the stack into two threat classes.
Retroactive — Harvest Now, Decrypt Later. An adversary records ciphertext today and decrypts it once a cryptographically relevant quantum computer exists. This applies only to confidentiality primitives. Its defining property is that the clock is already running: every day a confidentiality surface remains classical, the corpus of permanently exposed traffic grows, and no later migration undoes it. For these surfaces the deployment date is the security parameter.
Real-time forgery. An adversary with a CRQC at protocol-execution time forges signatures and proofs. This applies to integrity and authentication primitives. Recording today’s signatures and proofs confers no future advantage — a signature is a claim verified at receipt, and its purpose is complete at that moment. These migrations face a hard deadline before CRQC availability, but no accumulating penalty for taking the time to do them properly — with one caveat: a node syncing the chain after a CRQC exists re-verifies proofs whose soundness has lapsed. This is the post-CRQC bootstrap problem excluded from scope above; its standard mitigation, committing to the pre-migration chain under hash-based checkpoints before a CRQC arrives, is part of why the deadline is hard.
One nuance is worth stating explicitly, because it does more work than any other single observation in the analysis. A zero-knowledge proof’s threat class is set by its role in the protocol, not by the proof system. Groth16 is perfectly zero-knowledge: a recorded proof reveals nothing about its witness even to a computationally unbounded adversary. Where the stack uses proofs to protect confidentiality — the Proof of Leadership and the Proof of Claim both hide which note won and which block a reward stems from, and that unlinkability is a protected asset of Cryptarchia’s private proof-of-stake — the hiding is information-theoretic and the surrounding commitments are hash-based. Leader privacy therefore survives a quantum adversary both retroactively and prospectively. The only quantum-exposed property of every proof in the stack is soundness, which is computational. That is what places all four circuits, without exception, in the real-time-forgery class.
It also produces a requirement that must not be lost in the proof-system migration: the replacement must preserve zero-knowledge, not merely soundness. A merely-sound successor would keep the ledger unforgeable while silently giving up leader privacy.