The parent document says what the post-quantum transition costs. This note says what we should do about it.
Three things. The first has to happen before the chain starts. The second and third can come later, but the second decides how large blocks can be, so it is worth planning early.
Do this before genesis. It costs almost nothing now and it removes the entire migration later.
A public key is a hash of a secret key. Today that hash works over the field Groth16 uses. The post-quantum prover works over a much smaller field, and computing today's hash inside it is very expensive: that single mismatch is what creates the emulated circuits, the phone memory problem, the two proof styles and the whole migration story in the cost analysis.
The fix is to have every note commit to both keys from the start.
pk_classic, derived as today, Poseidon2 over the BN254 field.pk_pq, derived with Rescue over Goldilocks, which is what the post-quantum prover speaks natively.Both are mandatory in the note. The note identifier is derived from both.
Before the switch, spending proves the link to pk_classic, exactly as Groth16 does today. Nothing gets more expensive.
After the switch, spending proves the link to pk_pq natively. Nodes rebuild the note identifiers and the ledger trees from pk_pq alone, and pk_classic is dropped (if desired we could support pk_classic during a transition period to guarantee no funds loss but enforce every new created note to only have a single pk_pq key).
That is it. No emulated circuit, no two signature circuits, no incentive scheme to nudge people across, no migration period. The note carries the bridge, so the proof never has to.
A wallet must check both keys are its own when it receives a note.
The sender writes the keys into the output. A careless or hostile sender can write a correct pk_classic and a pk_pq the recipient does not control. That note is perfectly spendable today and dead at the switch, because nobody can produce a proof for a key they do not hold.
Checking at receipt makes this a non-problem. The recipient sees it immediately, and can refuse the payment or move the note straight away. Skipping the check means finding out years later, when the funds are gone.
So this is not a recommendation, it is a requirement on any wallet that wants to be safe. The recommended derivation should be published alongside the note format, so that all normal software derives pk_pq the same way and the check is trivial.
One thing to decide. Whether to keep the emulated signature circuit available for a while after the switch, purely as a rescue path for notes whose pk_pq is wrong. It should never be needed. It is the only way to move such a note. Building an expensive circuit for a case that should not happen is a real question and not an obvious yes.
Post-quantum proofs are large, and none of them can be thrown away. A signature proof establishes that the ledger was built correctly, and the ledger decides who produces blocks, so a node that discards one can no longer validate the chain on its own.