# The Two-Moiré Argument (Chris, 2026-08-14)

<!-- ledger
id: Q-two-moire
status: ANSWERED
todo: none
question: Do the Grain and the Scour, generated by disjoint prime alphabets, force twins to be infinite?
verdict: The provable content is exact CRT factorization: over the JOINT period the miss count is guaranteed positive, so aggregate alignment is arithmetically impossible, but that is an aggregate statement over a period vastly wider than the zone and does not reach TPC; the nearest unclaimed theorem is a two-class Jacobsthal upper bound at ANY exponent, and exponent < 2 would prove TPC outright.
-->

The most framework-native formulation of the twin prime question yet. Recorded
with its assertion-by-assertion verdict.

## The argument

1. The twin slots of Tₓ form a known pattern (**the Grain**) — Copying Theorem.
2. Their spacing distribution is fixed and knowable (Twin Prime Grain).
3. Folding preserves grain values (copies) and coarsens only at kills (merges).
4. Only primes in (x, √|Tₓ|] can strike slots inside the tile (**the Scour** —
   the tile's own crystallized holes, re-scaled: ⋃ q × holes).
5. The Scour is itself a knowable moiré with fixed misses.

**Assertion:** Grain and Scour are generated by DISJOINT prime alphabets, so
they cannot align; therefore twins are infinite.

## Verdict

- **Provable (theorem):** the disjoint alphabets give EXACT CRT independence —
  the joint distribution factorizes; the Scour strikes slots at exactly rate
  2/q, with zero preference. Over the JOINT period (width |Tₓ|·∏ scour-q),
  the miss count is guaranteed: census · ∏(q−2) survivor classes > 0.
  Aggregate alignment is arithmetically impossible. This factorization is the
  engine of the Variance Theorem and of every Hardy–Littlewood constant we
  have verified (E(P) to 0.6%, HL window counts to ±1σ, etc.).
- **The gap (the wall, in two-moiré form):** the guaranteed misses are
  residues of the joint tile — width ~10⁷³ already at x=13 — while the
  region of interest [0, |Tₓ|) is a ~10⁻⁶⁹ sliver of it. Independence gives
  the right EXPECTED miss count in the sliver (predicted 456 = measured 456)
  but no guarantee; total local alignment inside one sliver cannot be excluded
  by counting (parity). The Euclid rescue fails for pairs (both members must
  be prime, not merely have new factors).
- **Net:** "misaligned on average" is a theorem; "misaligned in every window"
  IS the Twin Prime Conjecture. The conjecture, restated in final
  primeoire-native form:

> **The Scour never achieves perfect local alignment with the Grain.**

## The mod-30 refinement and house-blindness (2026-08-14, later same day)

- **The tick:** the @5 family is exactly {r ≡ 11 or 17 (mod 30)} — a rigid
  two-tooth comb carrying a frozen 2/3 of every census.
- **House-blindness (provable, CRT + measured):** every remover is coprime to
  30, so the Scour strikes the three houses in exact proportion — measured on
  the T₁₃ ledger: kill rates 68.7% / 69.3% / 69.9%, survivors 155/152/149.
  The Scour cannot preferentially hunt any house.
- **Perfect-overlap question, three scopes:** joint period — impossible
  (theorem); all integers — impossible (census > 0); inside each tile window
  for all large x — this IS TPC in mod-30 costume.

## The covering-systems connection (the new arena)

Asking "can classes {0, −2 mod q}, primes q > x only, cover a stretch of the
comb?" lands in the covering-systems literature:

- **Hough 2015** (Erdős minimum-modulus problem): no covering system of ℤ has
  all moduli large. The Scour is exactly an attempted all-large-moduli
  covering — the infinite version of the overlap question is settled (also
  trivial for us via census > 0, but the kinship is structural).
- **The finite version is the Gap Reformulation's home turf:** covering an
  interval with one class per small prime = the Erdős–Rankin large-gap
  construction; best known coverage (FGKMT 2018) ~ y·ln y·(small factors) —
  the adversary's best known weapons fall short of the zone's p² by ~p/ln p.
  A TPC failure requires a covering phenomenon of a fundamentally new kind.
- **The nearest unclaimed theorem:** the two-classes-per-prime Jacobsthal
  upper bound is open at ANY exponent (audit, re-run in A144311's own wording rather than in ours: no published analog of
  Iwaniec's g(q) ≪ ln²q for two classes; Ziller–Morack's bound is
  conjectural). Any upper bound at all would be new; exponent < 2 would
  prove TPC outright (via crystallization + Gap Reformulation). The precise
  technical question: where exactly does Iwaniec's proof break for two
  omitted classes per prime?

## Why this formulation still earns its place

It is the cleanest intuition for why every expert believes the conjecture
(disjoint alphabets cannot conspire), it makes the provable content precise
(exact factorization — no hand-waving), and it locates the entire difficulty
in one image: a 10⁻⁶⁹ sliver of a guaranteed pattern. Paper chapter 4 closer.