// ============================================================================ // THE TWIN-GAP RECORD LADDER ADOPTED AS TRUSTED DATA — A113274 + A113275, // ALL 82 PUBLISHED RECORDS, AND THE 0.76 ln^3 p GUARD READ RECORD-EXACT // // Question: ZONE-POSTULATE's guard against the postulate being tight — the // first-twin distance grows like polylog while the window grows like p^2, // with Kourbatov's published ceiling "maximal gaps between twin primes are // less than 0.76 log^3 p" — was argued on eleven in-house decades (largest // extreme 8,042 at p = 65,095,731,749) plus four decade-slopes quoted to // 10^15. The full record ladder is published. What does the guard read like // on every known record, record-exact instead of slope-summarised? // // Provenance of the trusted series, read at the OEIS entries 2026-08-20: // A113274 record gaps between twin primes (gap = lesser to lesser). // B-file: Martin Raab, n = 1..82; terms up to a(72) from Alexei // Kourbatov, a(73)-a(75) from Tomas Oliveira e Silva, a(76)-a(82) // from Raab. Largest record 35,640. // A113275 lesser of the twin pair STARTING each record gap (same n). // B-file: Raab, first 75 terms from Max Alekseyev. // Convention, checked three ways below: the gap is measured lesser-to-lesser // (a(1) = 2 from 3,5 -> 5,7); a(n) = 0 mod 6 for n > 1 (entry comment); and // our own eleven-decade run's extreme is record n = 41 — same start prime, // gap 8,040 = our 8,042 minus 2, because ZONE-POSTULATE measured to the // UPPER member of the arriving pair. A +-2 convention shift reported as an // extension would be worse than a miss; it is asserted, not assumed. // // Honest doubt: the b-file tails are single-witness per range (Kourbatov / // Oliveira e Silva / Raab), trusted under the series rule of 2026-08-20; // in-house verification is a paper-phase deliverable. The first 21 records // are recomputed here from a direct sieve (an in-house custody anchor, the // same role the 14-term overlap plays in a144311-full-ladder.js). // // Start primes a(76)..a(82) exceed 2^53, so both series are BigInt; logs go // through Number() only inside Math.log, where the rounding is ~1e-16 // relative and far under the 4 figures quoted. Runtime: ~2 s (the sieve). // ============================================================================ 'use strict'; // A113274 verbatim (record gaps, lesser to lesser), b-file 2026-08-20: const GAP = [ 2n, 6n, 12n, 18n, 30n, 36n, 72n, 150n, 168n, 210n, 282n, 372n, 498n, 630n, 924n, 930n, 1008n, 1452n, 1512n, 1530n, 1722n, 1902n, 2190n, 2256n, 2832n, 2868n, 3012n, 3102n, 3180n, 3480n, 3804n, 4770n, 5292n, 6030n, 6282n, 6474n, 6552n, 6648n, 7050n, 7980n, 8040n, 8994n, 9312n, 9318n, 10200n, 10338n, 10668n, 10710n, 11388n, 11982n, 12138n, 12288n, 12630n, 13050n, 14262n, 14436n, 14952n, 15396n, 15720n, 16362n, 16422n, 16590n, 16896n, 17082n, 18384n, 19746n, 19992n, 20532n, 21930n, 22548n, 23358n, 23382n, 25230n, 26268n, 28842n, 31302n, 31482n, 31512n, 31920n, 33042n, 34692n, 35640n ]; // A113275 verbatim (lesser twin starting each record gap): const START = [ 3n, 5n, 17n, 41n, 71n, 311n, 347n, 659n, 2381n, 5879n, 13397n, 18539n, 24419n, 62297n, 187907n, 687521n, 688451n, 850349n, 2868959n, 4869911n, 9923987n, 14656517n, 17382479n, 30752231n, 32822369n, 96894041n, 136283429n, 234966929n, 248641037n, 255949949n, 390817727n, 698542487n, 2466641069n, 4289385521n, 19181736269n, 24215097497n, 24857578817n, 40253418059n, 42441715487n, 43725662621n, 65095731749n, 134037421667n, 198311685749n, 223093059731n, 353503437239n, 484797803249n, 638432376191n, 784468515221n, 794623899269n, 1246446371789n, 1344856591289n, 1496875686461n, 2156652267611n, 2435613754109n, 4491437003327n, 13104143169251n, 14437327538267n, 18306891187511n, 18853633225211n, 23275487664899n, 23634280586867n, 38533601831027n, 43697538391391n, 56484333976919n, 74668675816277n, 116741875898981n, 136391104728629n, 221346439666109n, 353971046703347n, 450811253543219n, 742914612256169n, 1121784847637957n, 1149418981410179n, 2543288406389231n, 2797282815481499n, 12914226879316517n, 16155559543324757n, 37338553629118097n, 37962553054417547n, 38617975949216087n, 52000545890760149n, 70478530884377381n ]; // --- transcription and convention guards ------------------------------------ if (GAP.length !== 82 || START.length !== 82) throw new Error('term counts'); for (let i = 1; i < 82; i++) { if (GAP[i] % 6n !== 0n) throw new Error('a(' + (i + 1) + ') not 0 mod 6'); if (GAP[i] <= GAP[i - 1]) throw new Error('gaps not increasing at n=' + (i + 1)); if (START[i] <= START[i - 1]) throw new Error('starts not increasing at n=' + (i + 1)); } // the in-house eleven-decade extreme (ZONE-POSTULATE.md): p = 65,095,731,749, // distance to the next twin's UPPER member 8,042. Must be record n = 41. if (START[40] !== 65095731749n || GAP[40] !== 8040n || GAP[40] + 2n !== 8042n) throw new Error('in-house extreme does not match record 41 — convention broken'); // recompute the first 21 records from a direct sieve (starts < 1.2e7): { const N = 15000000; const s = new Uint8Array(N + 1); for (let i = 2; i * i <= N; i++) if (!s[i]) for (let j = i * i; j <= N; j += i) s[j] = 1; const lesser = []; for (let i = 3; i + 2 <= N; i += 2) if (!s[i] && !s[i + 2]) lesser.push(i); let rec = 0, k = 0; for (let i = 1; i < lesser.length; i++) { const g = lesser[i] - lesser[i - 1]; if (g > rec) { rec = g; if (k < 21) { if (BigInt(g) !== GAP[k] || BigInt(lesser[i - 1]) !== START[k]) throw new Error('sieve record ' + (k + 1) + ' mismatch: gap ' + g + ' at ' + lesser[i - 1]); k++; } } } if (k !== 21) throw new Error('sieve found only ' + k + ' of 21 records'); } // --- the guard, record by record -------------------------------------------- // Kourbatov's ceiling is evaluated at the prime ENDING the gap; here that is // the lesser of the arriving pair, p_end = START + GAP. const rows = []; for (let i = 0; i < 82; i++) { const pEnd = START[i] + GAP[i]; const ln = Math.log(Number(pEnd)); const ceil = 0.76 * ln * ln * ln; rows.push({ n: i + 1, gap: GAP[i], pEnd, load: Number(GAP[i]) / ceil, margin: Number(pEnd) / ceil }); } console.log('THE 82 TWIN-GAP RECORDS AGAINST THE PUBLISHED GUARD 0.76 ln^3 p'); console.log('guards: mod 6 81/81, strictly increasing 82/82, in-house extreme = record 41,'); console.log(' first 21 records recomputed by direct sieve: EXACT'); console.log(''); console.log(' n gap p_end gap/(0.76 ln^3 p)'); for (const r of rows) { if (r.n <= 3 || r.n % 10 === 0 || r.n >= 73 || r.n === 41 || r.n === 75) { console.log(String(r.n).padStart(4) + ' ' + String(r.gap).padStart(10) + ' ' + String(r.pEnd).padStart(21) + ' ' + r.load.toFixed(4).padStart(12)); } } console.log(' (all 82 checked; table shows every tenth, the head, the in-house'); console.log(' anchor n = 41, and the tail)'); console.log(''); const worst = rows.reduce((a, b) => (b.load > a.load ? b : a)); console.log('records above the ceiling: ' + rows.filter(r => r.load >= 1).length + ' of 82'); console.log('worst load: ' + worst.load.toFixed(4) + ' at n = ' + worst.n + ' (gap ' + worst.gap + ', p_end ' + worst.pEnd + ')'); const last = rows[81]; console.log('last record: gap ' + last.gap + ' at p_end ' + last.pEnd + ', load ' + last.load.toFixed(4)); console.log('window over guard at the last record: p^2/(0.76 ln^3 p) = ' + (Number(last.pEnd) * last.margin).toExponential(2) + ' (the margin that runs away)'); const loads10 = rows.filter(r => r.load > worst.load - 0.1).length; console.log('records within 0.1 of the worst load: ' + loads10 + ' of 82'); // ============================================================================ // OUTPUT — EMBEDDED, do not hand-edit. Regenerate: // node research/qc/embed.js research/a113274-gap-records.js // invocation: node research/a113274-gap-records.js // code-sha256: ce36ef40ad2faa0871565627ecca4478eb9cc04e35a81b552bd00f0fcb886eab // out-sha256: 008ad4855d217fc2476843c5b3a72b3dbea775e4dc0ceba49682b3c4eccd5b08 // streams: stdout // node: v22.21.0 // embedded: 2026-08-20 // elapsed: 0.1 s // ============================================================================ // THE 82 TWIN-GAP RECORDS AGAINST THE PUBLISHED GUARD 0.76 ln^3 p // guards: mod 6 81/81, strictly increasing 82/82, in-house extreme = record 41, // first 21 records recomputed by direct sieve: EXACT // // n gap p_end gap/(0.76 ln^3 p) // 1 2 5 0.6312 // 2 6 11 0.5726 // 3 12 29 0.4135 // 10 210 6089 0.4176 // 20 1530 4871441 0.5513 // 30 3480 255953429 0.6310 // 40 7980 43725670601 0.7139 // 41 8040 65095739789 0.6853 // 50 11982 1246446383771 0.7298 // 60 16362 23275487681261 0.7384 // 70 22548 450811253565767 0.7723 // 73 25230 1149418981435409 0.7960 // 74 26268 2543288406415499 0.7744 // 75 28842 2797282815510341 0.8434 // 76 31302 12914226879347819 0.8067 // 77 31482 16155559543356239 0.7969 // 78 31512 37338553629149609 0.7462 // 79 31920 37962553054449467 0.7549 // 80 33042 38617975949249129 0.7804 // 81 34692 52000545890794841 0.8005 // 82 35640 70478530884413021 0.8032 // (all 82 checked; table shows every tenth, the head, the in-house // anchor n = 41, and the tail) // // records above the ceiling: 0 of 82 // worst load: 0.8434 at n = 75 (gap 28842, p_end 2797282815510341) // last record: gap 35640 at p_end 70478530884413021, load 0.8032 // window over guard at the last record: p^2/(0.76 ln^3 p) = 1.12e+29 (the margin that runs away) // records within 0.1 of the worst load: 18 of 82 // ============================================================================ // READINGS // // 1. THE TRUST'S GATES PASSED, THREE OF THEM. The entry's own invariant (all // records past the first are 0 mod 6) holds 81/81; both series are // strictly increasing as a record ladder must be; the first 21 records // are recomputed exactly by a direct sieve; and the eleven-decade run's // extreme (ZONE-POSTULATE.md) is record n = 41 with gap 8,040 — our // in-house figure differs by exactly 2, the measurement convention // (lesser-to-lesser here, lesser-to-upper there), caught at the record // instead of shipped as a phantom extension. // 2. THE GUARD HOLDS AT EVERY RECORD EVER PUBLISHED: 0 of 82 reach // Kourbatov's ceiling 0.76 ln^3 p. The worst load is 0.8434 (record 75, // gap 28,842) and the last record's is 0.8032. The guard check moves // from eleven in-house decades summarised by four slopes to record-exact // coverage six decades further, and the postulate-side margin at the // last record is p^2/(0.76 ln^3 p) = 1.12e+29. // 3. THE LOAD CREEPS, WHICH IS KOURBATOV'S OWN READING, NOT A SURPRISE. 18 // of 82 records sit within 0.1 of the worst load, all in the ladder's // upper half, and his two formulas are asymptotically equal, so the load // should drift toward 1 from below without crossing. Record-exact data // shows exactly that shape: a slow rise with local dips, never a breach. // 4. WHAT THIS RETIRES: ZONE-POSTULATE's "nothing in eleven decades hints at // the postulate being tight" was true and under-sold — nothing in the // entire published record ladder does either, and the sentence there now // cites this artifact instead of stopping at our own run. // 5. LIMITS. Record ranges are single-witness per contributor (Kourbatov to // a(72), Oliveira e Silva to a(75), Raab beyond); trusted under the // series rule of 2026-08-20, in-house verification a paper-phase // deliverable. The ceiling itself is a heuristic (HL-derived, no proof — // his own framing), so this is a guard on a guard, not a theorem; and a // record ladder bounds FIRST-twin distances only through its records, so // the postulate-side reading stays an argument from margin, as before. // ============================================================================