# The lens: why the tile is the right object *(Written 2026-08-16 at Chris's request, to record why this framing has power before the details bury it. Canonical alias for the tile, stated once: the primorial wheel mod p#. Nothing here is claimed as new mathematics; the objects are classical. What the note records is why they are the right things to look at, and what they do and do not buy.)* **Scope.** This document owns the framing: why the tile is the right object, the exact recursion, the CRT covering identity, the one-number reformulations, and the triage rule of §5, which is the corpus's canonical statement of it. It does not own the state of any object (`G2-STATE.md`), the target (`ZONE-POSTULATE.md`), the vocabulary (`GLOSSARY.md`) or the work queue (`../TODO.md`). ## 1. The central property: completeness Fix a prime p and let W = p# be the product of all primes up to p. The tile T_p is the residue pattern mod W: its **holes** are the residues coprime to W, its **twin slots** the positions r where r and r+2 are both holes. **Completeness (PROVEN, elementary).** Every twin prime pair above p appears as a twin slot of T_p. *Proof.* Let (q, q+2) be twin primes with q > p. Both are primes exceeding p, so neither is divisible by any prime up to p, so both are coprime to W. Hence r = q mod W is a twin slot. ∎ That is the whole engine, and it is worth stating baldly: **the tile misses nothing.** It is a finite, periodic, exactly computable pattern that contains every twin prime there will ever be. There is no tail, no error term, no sampling. The twin primes are a subset of a set we can write down completely. The converse holds in a region, which is what makes the containment useful. **Crystallization (PROVEN, elementary).** Every hole h of T_p with 1 < h < p′², where p′ is the least prime above p, is a genuine prime; every twin slot with opener greater than 1 and both members below p′² is a genuine twin prime pair. *Proof.* A hole h > 1 is coprime to every prime up to p, so if composite its least prime factor exceeds p, forcing h ≥ p′². ∎ So inside the **zone**, containment becomes equality: the pattern is not a superset of the truth, it IS the truth. Above the zone the pattern is a superset that later folds prune. The proof of crystallization never uses primality of p, so the same statement holds with any u in place of p. That is what lets the frontier be written in closed form, u², with no next-prime lookup, and it is the basis of the u-frame (research/U-FRAME.md). ## 2. The recursion is exact Folding by the next prime is not an approximation. It is an exact rewriting. **The fold (PROVEN).** T_{p'} consists of p′ copies of T_p laid end to end, with p′ striking the residues 0 and −2 in every copy. Each existing slot keeps exactly p′ − 2 of its p′ copies, which gives the census law D_{p'} = (p′ − 2)·D_p. **The residue-deletion form (PROVEN, verified at six folds).** Since gcd(W, p′) = 1, as k runs over the p′ copies the shift kW mod p′ runs over every residue exactly once. Writing w = W mod p′ and r_i = s_i mod p′, copy k deletes exactly the slots whose residue lies in the 2-set {−kw, −kw−2}. So a fold is p′ residue-class deletions applied to one fixed pattern, and G₂(new) = max over the p′ 2-sets of the damage that deletion does. ## 3. What CRT permits across folds **PROVEN.** Every alignment for a new prime occurs among the copies. Moreover, for any choices `a_p` at every prime in the tile, CRT supplies a phase `s` with `−s ≡ a_p (mod p)` simultaneously. The forbidden positions in a window translated by s are exactly `{a_p, a_p−2}` for each p. Thus adverse translated pairs can occur together. Independence of the coordinates under a uniform random phase does not exclude any joint configuration in a maximum over phases. Consequently `G₂(P)−1` is precisely the maximum covered interval when one may translate each prime's pair freely while keeping separation 2. This is the covering identity in `two-class-lower-bounds.md` §1, checked exhaustively at T7 by `review-0905-validation.js`. The paired Jacobsthal function in OEIS A288815 also varies the common even offset. Its larger values do not establish protection against translations in our fixed-offset object. Record relocation at measured folds is finite data; it supplies no theorem preventing damage from accumulating at some phase. Any advantage at the integer origin needs a separate anchored estimate. ## 4. What the lens buys **Many local residue identities are elementary.** The census ∏(q−2), the Copying Theorem, the Seam Lemma, the palindrome, the birth cohorts, count(6) = ∏(q−4), the never-self-strike classes, the Inertness Lemma: all of these fall out of the CRT product in a line or two. The lens produces them constantly and cheaply. **Exact computation at any level.** Because the object is finite and periodic, quantities that are asymptotic elsewhere are integers here. The census, the gap word, the variance, the anchored survivor count, the twin Jacobsthal are all exactly computable, and have been computed: β at ten levels out to W = 3.04e14, the ensemble variance at nine, G₂ at twelve. **One-number reformulations.** The lens keeps compressing the conjecture into the positivity or boundedness of a single computable quantity. Five so far, all equivalent to or stronger than TPC. The u-frame gets two rows because its two forms have different statuses, and two rows are already priced out as routes: | reformulation | the one number | status as a route | file | |---|---|---|---| | anchored bias | liminf β(x) > 0 | open, Hardy-Littlewood strength | paper/anchored-note.md | | Gap Reformulation | two-class Jacobsthal exponent < 2 | open sufficient target; secondary to the long-interval benchmark | paper/beta2-note.md | | Zone Postulate | first twin slot below u² | open, the stated focus | research/ZONE-POSTULATE.md | | Pane Bound | removals below slots in (n², (n+2)²) | CLOSED, strictly harder than the zone | research/GLOSSARY.md | | u-frame, multiplier form | ln c(p) ≤ 2 ln p / p per fold | CLOSED, the rate has no slack | research/U-FRAME.md | | u-frame, copy-theorem form | a bound on the residue-deleted maxsum | open, and the branch's one live entry point (TODO 0c) | research/U-FRAME.md | The two CLOSED rows carry their own verdict and are worth reading together. The Pane Bound removes the zone's redundancy and pays for it: in the repo's own currency it reads "4.2665 proven, 1 needed" against the zone's "4.2665 proven, 2 needed", and 1 is half the parity floor. Removing the redundancy moved the target further away, not nearer. The u-frame's per-fold requirement is sharp rather than generous. It comes from differentiating the partial-sum condition Σ_{p≤x} ln c(p) ≤ 2 ln x − ln 12, and the multiplier the tile actually runs is measured at 0.77 to 2.14 times that rate on our own ladder and at mean 1.06 on Ziller and Morack's dominating sequence. What is left is not a rate advantage but a total lifetime slack measured at 0.88 to 1.19 nats, summed over every fold from here to infinity. Both are correct reformulations. Neither is a way in. That compression is the lens's real product. It does not make the conjecture easier, but it makes it *sayable* in terms of finite objects, which is the precondition for aiming at it. ## 5. What the lens does not buy, and the rule that keeps us honest **The triage rule.** Identify the statistic, the scale, the quantifier and which arithmetic estimates the argument actually uses. Local CRT counts are often elementary and classical. An anchored survivor estimate can be hard; expressing it in residue notation does not make it easy or prove it impossible. Exact inclusion–exclusion retains all the information, but its signed remainder still requires control on the relevant intervals. Completeness says the tile contains every twin pair above the sieve depth. It does not provide a lower bound in the chosen window. An almost-all-phase estimate does not automatically apply at the integer origin. Any argument that uses equidistribution there must state whether it has proved, imported, or assumed the required error bound. **A working test.** Trace the step that produces positive occupancy. Check that its hypotheses, interval length and errors match the actual conclusion. A finite measurement cannot supply an unproved asymptotic input. The test has already caught a published programme. Holt's Legendre result (arXiv 2603.25915 Theorem 3.3) rests on his Conjecture 2.1, "approximate uniformity", stated as a conjecture supported by samples. That is the failure mode above, reached from inside two decades of work on these same objects. The rule named the shape before we found the example, which is the reason to run it on our own arguments first rather than last. **And the ceiling.** Every reformulation in the table above is equivalent to or stronger than TPC, so each contains something at least TPC-hard. No reformulation here has removed that difficulty. This records the status of the existing arguments; it does not exclude an independently proved estimate in these coordinates. ## 6. Positioning A new framework and vocabulary over classical sieve-theoretic objects: a new lens. The tile is the primorial wheel, the census is Schemmel's totient at the primorial, the fold is the wheel recursion, and the Zone Postulate route is Ziller and Morack's from 2017. **The largest overlap is with Fred B. Holt, with Helgi Rudd on the earlier papers, and it covers most of this note.** Their programme has run since 2007, about fifteen manuscripts, and it independently occupies the frame: the tile is his cycle of gaps G(p#) among the generators of Z mod p#; §2's fold is his R1/R2/R3 and Lemma 2.1; the kills are his fusions; the Copying and Redundancy Lemmas are his Theorem 2.3; the census ∏(q−2) is his N2(p#); §1's zone is his interval of survival Δ-H(p_k) = [p_k², p_{k+1}²] and the crystallization frontier is his horizon of survival. The histogram transfer operator is his and Rudd's from 2014, eigenstructure included. See `research/PRIOR-ART.md` for the correspondence table and the citations. We reached these from the corpus and from first principles without knowing the programme existed, several of the papers not being in print until 2025 and 2026, so independent arrival is the credential. Priority is theirs and we cite it. What is ours is the vocabulary, the exact computations, the honest boundary, the compressions of §4, the two-class form of the merge structure and its application to a maximum gap, and the triage rule of §5. --- *This document states current understanding. Superseded claims, retired numbers and the reasons they changed are in [history/CHANGELOG.md](history/CHANGELOG.md), indexed by document.*