--- work/job2987/attack-kstar-01.served.md 2026-09-24 23:17:14.708282000 +0200 +++ artifacts/job2987/attack-kstar-01.md 2026-09-24 23:24:44.493141500 +0200 @@ -1,11 +1,11 @@ -# K* extended five steps: the drift is structural, the certificate reaches 18 of 19.2455 at the chain step itself, and the HL first moment becomes a period-free certificate engine +# K* extended six steps: the drift is structural, the certificate reaches 18 of 19.2455 at the chain step itself, and the HL first moment becomes a period-free certificate engine *2026-08-21. Producer: `research/attack-kstar-01.js` (345 s, formal embed @@ -26,11 +26,11 @@ ## 0. The verdict, up front -> **The alarm is structural, and it lands early. K* is now enumerated at 16 +> **The alarm is structural, and it lands early. K* is now enumerated at 17 > doubling steps (was 11): three new full period walks (13#→29#, 13#→31#, -> 17#→31#) and two period-free certificates past every scannable level -> (19#→37#, 19#→41#). The raw drift slope steepens from 0.6881 ± 0.1328 to -> 0.8184 ± 0.0908, and the certificate touches K*+1 = 18 — 93.5% of +> 17#→31#) and three period-free certificates past every scannable level +> (19#→37#, 19#→41#, 19#→43#). The raw drift slope steepens from 0.6881 ± +> 0.1328 to 0.8184 ± 0.0908, and the certificate touches K*+1 = 18 — 93.5% of > 2^β₂ = 19.2455 — at the base-2 chain step s = 16 itself, the sup row. > The instrument that did it: the kill-run count N_k is EXACTLY an > alternating sum of Hardy–Littlewood local products (W1's ν-case table one @@ -41,7 +41,7 @@ > cells), and the HL growth verdict is NEAR-LINEAR: the bridging-certificate > route to an all-s C₂ < 19.2455 is dead — under the model it fails at the > very next chain step 31#→61#, and no fit is needed to see cert = 18 at -> P(2s) = 31. C₂ itself stays flat (3.22–5.27 across the five new steps); +> P(2s) = 31. C₂ itself stays flat (3.22–5.27 across the six new steps); > what dies is the certificate, not the inequality.** ## 1. (a) The extended ladder @@ -50,7 +50,7 @@ re-deriving the exact ladder from scratch as its custody gate — **G2(29#) = 258 @ 1205437109 (×2)** and **G2(31#) = 348 @ 8813641451 (×4)**, digit-for-digit including least argmax and multiplicity, alive counts -= Π(p−2) — plus two certificate-only steps (§3). The fold-recursion path was += Π(p−2) — plus three certificate-only steps (§3). The fold-recursion path was also pushed from 13# to 23# (D₂₃ = 7,952,175), so ladder rows 17, 19, 23 are now re-derived by BOTH paths. **[VERIFIED]** @@ -61,12 +61,14 @@ | 17#→31# | 17, 18 | 13 | 14 | 29/9 = 3.2222 | 4.34 | scan | | 19#→37# | 19, 20 | 13 | 14 | 528/150 = 3.5200 | 3.98 | I–E, period-free | | 19#→41# | 21 | 16 | 17 | 546/150 = 3.6400 | 4.67 | I–E, period-free | +| 19#→43# | 22, 23 | 20 | 21 | 618/150 = 4.1200 | 5.10 | pruned covering DP (kstar_dfs.py, #1144), period-free | -Full ladder (K*+1, 16 steps): 3 2 5 3 4 4 6 9 7 11 9 11 18 14 14 17. -All five new C₂ comparisons are custody-exact (both G₂ terms among the 14 -exact ladder terms). The certificate holds at every step and its slack -cert/C₂ **widens monotonically in scale**: 1.00 at the first three steps -(attack-doubling-01) to 4.67 at 19#→41#. **[VERIFIED]** +Full ladder (K*+1, 17 steps): 3 2 5 3 4 4 6 9 7 11 9 11 18 14 14 17 21. +All five of this pass's new C₂ comparisons are custody-exact (both G₂ terms +among the 14 exact ladder terms); the sixth, 19#→43#'s, is exact on the same +footing by its own report (#1144). The certificate holds at every step and +its slack cert/C₂ **widens monotonically in scale**: 1.00 at the first three +steps (attack-doubling-01) to 5.10 at 19#→43#. **[VERIFIED]** ## 2. (b) The HL route, made exact — and the growth verdict @@ -110,7 +112,7 @@ certificates in). The slope went UP with the new data. **[MEASURED]** - The decisive instrument I_eff = ln(D·M)/K*: a log-type K* needs I_eff growing like P/ln P (roughly ×2 from P(2s) = 23 to 43); measured I_eff - stays in **[0.55, 2.56]** across all 16 steps, raw trend 0.3676 ± 0.1400 + stays in **[0.55, 2.56]** across the same 16 steps, raw trend 0.3676 ± 0.1400 per ln P — bounded-looking, mildly rising, nowhere near doubling. Since ln(D·M) = θ(P(2s)) − ln(mean slot gap) grows linearly in P(2s), bounded I_eff forces K* to grow near-linearly. **[MEASURED + INFERRED]** @@ -122,7 +124,7 @@ ~22–28 — every value above 19.2455 (illustration, not a claim; the producer prints the band). **[MEASURED the inputs; INFERRED the band]** -What this does NOT say: C₂ shows no drift (the five new values sit in +What this does NOT say: C₂ shows no drift (the six new values sit in [3.22, 5.27], inside the known range, sup still 5.2727 at s = 16). The inequality's truth is untouched; the *certificate quantity* grows. The route "prove C₂ ≤ K*+1, bound K*" cannot give any all-s C₂ < 19.2455, @@ -156,8 +158,10 @@ problem again, the same self-similarity attack-doubling-01 §3 named. The bypass changes the cost of a finite certificate (period → tile), not the quantifier. **[INFERRED]** -- **Where the engine's own reach ends.** 19#→43# needs a peak pass of - ~2.5·10¹⁰ subsets (priced from the measured prune profile), 23#→43# +- **Where the engine's own reach ends.** 19#→43# was reached by the pruned + covering search in minutes — K* = 20 over all 378,675 starts in 329 s + (`kstar_dfs.py`, #1144; rerun in review 2985) — where this engine's own + price for the same step is a peak pass of ~2.5·10¹⁰ subsets. 23#→43# ~1–2·10¹⁰ with D₂₃ = 7.95·10⁶ shapes — an hour-class JS run each; base 29 (D₂₉ = 2.15·10⁸) is out entirely. Not run. **[MEASURED the profile; the prices are estimates]** @@ -214,8 +218,8 @@ ## 6. Trap grading -Nothing here derives an all-s constant; the five new certificates are -finite-level theorems (C₂ ≤ 11, 18, 14, 14, 17 at their steps) and imply +Nothing here derives an all-s constant; the six new certificates are +finite-level theorems (C₂ ≤ 11, 18, 14, 14, 17, 21 at their steps) and imply nothing about limsup. All new C₂ measurements sit in [3.22, 5.27] — the sup 5.2727 and the landing zone [5.2727, 19.2455) are unchanged, TPC through the slice stays excluded by the s = 16 datum. The slice remains @@ -224,13 +228,16 @@ ## 7. NOT REACHED - **No all-s bound on C₂ or on K*.** The near-linear verdict is a 16-point - measurement plus HL-model asymptotics, not a theorem; growth past - P(2s) = 41 is inferred, not enumerated. -- **19#→43#, 23#→43#, 23#→47#**: priced (§3), not run; base 29+ out of - reach for the I–E engine as written. -- **19#→41#'s K* = 16 rests on the I–E route alone** — the period 3.04·10¹⁴ - admits no verifying walk; confidence is the 14-step dual-engine - validation, not an independent measurement at that step. + measurement (the 19#→43# row is not in that fit) plus HL-model asymptotics, + not a theorem; growth past P(2s) = 43 is inferred, not enumerated. +- **23#→43#, 23#→47#**: priced (§3), not run; base 29+ out of reach for the + I–E engine as written. 19#→43# is no longer "not run": the pruned + covering search reached it in minutes (#1144, §3). +- **19#→41#'s K* = 16 no longer rests on the I–E route alone** — the + pruned covering search returns 16 at that step too (`kstar_dfs.py`, + #1144), a second, independent instrument; the period 3.04·10¹⁴ still + admits no verifying *walk*, so the confidence remains the 14-step + dual-engine validation plus that agreement, not a walk at that step. - **M1's residual (+2 at 13#→29#, −1 at 13#→31#)** has no third-moment model; the higher-order rigidity is measured, not derived. - **The two maximal 17-runs at 13#→31#** are predicted by Mirror-Sweep @@ -252,10 +259,14 @@ | `research/history/staging/w1-singular-series.md` | the HL ν-case identification this pass lifts from pairs to k-tuples | | `research/exact-g2-ladder.js`, `research/import-interp-01-bgt-defect.js` | the two ladders, parsed at run time, input-hashed by the embed | | `research/dhr-verification.md` row 1a | β₂ = 4.26645…, giving 2^β₂ = 19.2455, the ceiling the certificates are read against | +| `kstar_dfs.py` (served, return #1144) | the pruned covering search: the 19#→43# row's K* = 20, and its certificate cells N_20 = 4, N_21 = 0 | Reproduce: `node research/qc/embed.js --check --timeout 900 research/attack-kstar-01.js` (the fingerprint matches as of 2026-08-21; -the run re-walks both periods, ~6 minutes). +the run re-walks both periods, ~6 minutes). The 19#→43# row: `python +kstar_dfs.py 19 43` → `kstar: 20`, every start tested at length 21, so +N_21 = 0 — and `--count 20` → `N_k: 4` (return #1144). Its `seconds` field +is wall time and is the only field that differs between runs. ---