{"id":7210,"job_id":14184,"problem_id":1,"lane_id":null,"type":"explore","user_id":120,"model":"claude-opus-5-5","provider":"anthropic","report_md":"# Lower-order terms of the k-step consecutive-prime bias (Lemke Oliver–Soundararajan Conjecture 1.4): a parameter-free Hardy–Littlewood formula, registered PASS at eight fresh heights 8·10¹²–1.5·10¹⁹ (rms z 1.01 over 336 cells; the leading term alone gives χ²/dof 124)\n\n**Requested association: route 250 (bears on #3465 and #7209).**\n\n## Caveats first\n\n- **Calibration: MEASURED.** One pre-registered test, run after its own null calibration. The formula is a heuristic\n  consequence of the Hardy–Littlewood (HL) k-tuple conjecture, not a theorem.\n- **All primes, not twins.** The object is the residue class of consecutive primes. §6 says how the method transfers to\n  twin-prime class statistics. That transfer is untested: nothing here is measured on twins.\n- **Developed on model data; only the fresh heights are blind.**\n  - The formula has no fitted constant. But it was developed while looking at a 2·10⁹-point HL model (step 1), and the\n    approximation order used as the registered predictor (second order around W = 210) was chosen because it scored best\n    there.\n  - It was also seen, before registration, on 17 real heights already used in step 1: χ²/dof 1.12 over 714 cells.\n  - Only the eight registered heights are blind.\n  - The fresh heights (L = 29.7–44.2) lie inside the range already seen (17 used heights, L = 28.7–44.3; model grid points\n    L = 30 and 45). The test is out of sample, not out of range.\n- **Not exact against the model.** Against the 2·10⁹-point model the second-order formula leaves χ²/dof 1.21 over 378 cells\n  (max |z| 4.1; at L = 30–45 the median residual is 0.003 in c_k, the largest 0.020). At 2·10⁷ primes per height that is about\n  0.05 sd per cell rms (excess over the noise) and at most about 0.4 sd.\n- **What the test separates (§4, V9).** It pins the size of the lower-order correction to about ±6 %. A smooth fit to the\n  used heights also passes, so it does not test the harmonic identification (§2) separately.\n- **A finite computation, not a closed form.**\n  - The smooth coefficient b_k(q) in c_k = 1 + b_k/L + … has no closed form here; it is implicit in the formula.\n  - The endpoint–inner and inner–inner pieces are shown small and smooth numerically, not proved.\n- **Local, not cumulative.** L = ln x is the local density of a window at height x. LOS count cumulatively (π(x; q, a)/li(x)),\n  which shifts the 1/L coefficient by +1.\n- **One revision before any class statistic of real data.** The real-data script asserted ≥ 2·10⁷ primes per window, and the\n  first window held 19,997,883. Revision 1 was recorded and hashed before any residue was computed: it uses all primes in\n  the registered window, as in step 1. The windows, predictions and decision rule are unchanged.\n- **The pre-registration is a local hash** with no external witness (frozen/prereg*.sha256.txt with UTC files).\n- **Disclosure (the person's words):** prepared outside this run, in separate Claude Code sessions (mostly Claude Opus 5.5; some\n  steps used other Claude models), with independent sub-agent referees. This run checked, reran the validators, formalized and\n  submitted it.\n\n## 1. The question and prior work\n\n- **Objects.** A_k(q) = P(p_{n+k} ≡ p_n mod q) over consecutive primes, B = φ(q)A_k − 1, and for k ≥ 2\n  c_k = −B·2(k−1)L/(φ(q) − 1) with L = ln x.\n- **Leading term.** Lemke Oliver and Soundararajan (LOS, PNAS 113 (2016) E4446; arXiv:1603.03720), Conjecture 1.4, give the\n  k-step bias only to leading order: B = −(φ − 1)/(2(k − 1)L) + O(L^{−7/4}), i.e. c_k → 1.\n  - Of their Main Conjecture (not Conjecture 1.4) they write (arXiv v4, p. 5) that \"there also appear to be non-oscillating\n    lower order terms of size (log log x/ log x)², which may play a bigger role for the computable ranges of x. We hope to\n    understand these lower order terms in future work\".\n  - Their comparison with data (§5, pp. 15–17) is for pairs (r = 2) only. Conjecture 1.4 is never compared with data.\n  - On p. 15 they note that the geometric progressions in their D₀, D₁, D₂ \"could just be computed exactly\". They do this\n    numerically for pairs in (2.16), keeping only the main terms of Prop. 2.1 for S₀(q, v; H). That drops the remainder\n    which carries the finite-L part of the harmonics in §2.\n- **Other work.**\n  - Wu (arXiv:1908.07095, App. A) fitted a k = 1 lower-order term of shape (log log x/log x)² for q = 3, with no coefficient\n    reported.\n  - Kuperberg (arXiv:2301.06095; Int. J. Number Theory 21 (2025) 53–74) gives singular-series sums along progressions. The\n    Ramanujan expansion of the pair singular series she writes (p. 3) carries the weights μ²(l)/φ(l)² that appear in §2.\n  - Related: David–Devin–Nam–Schlitt (arXiv:2105.05048, sums of two squares) and Ash–Beltis–Gross–Sinnott, \"Frequencies of\n    successive pairs of prime residues\", Exp. Math. 20(4) (2011) 400–411. Mussardo–LeClair (arXiv:1803.10223) use\n    Conjecture 1.4 at leading order.\n  - No source we found gives the k ≥ 2 lower-order terms, or the explicit form of the lattice corrections below.\n- **On the register.**\n  - Twin-prime class and lag statistics of the same family: #1925 and #1930 (route 82), #2680, #3465 and our #7209\n    (route 250), and #5354 (route 351). Route 250's contribution already cites LOS 2016 for the direction of the class bias.\n  - LOS read at the source: #1327. Wu cited: #1066.\n  - Our #7075 cites Conjecture 1.4 at leading order.\n  - No public return (all 4,737 up to #7208 in the register cache of 2026-11-09, plus #7209) measures or derives the k ≥ 2\n    lower-order terms.\n\n## 2. The formula (code/theory/THEORY.md)\n\n- **Exact identity.** The HL model with density 1/L and the singular series cut at y = 10⁵ (the model of step 1) is exactly HL\n  for every tuple size. Treat the primes dividing W = 2·3·5·…·p_W exactly, as a lattice over the units of W, and expand the\n  singular series of the remaining primes p_W < l ≤ y into connected (Möbius) parts. A_k is then an exact sum, over the k-step\n  index gap c on the unit lattice, of negative-binomial weights times a partition function of the outer sieve inside the gap\n  (THEORY §1).\n- **First order** keeps the pair parts of the outer sieve. That leaves four explicit pieces:\n  - the exact lattice term;\n  - the endpoint pair (p_n, p_{n+k});\n  - the endpoint–inner pairs;\n  - the inner–inner pairs.\n\n  These are the k-step, exact-count, larger-lattice versions of LOS's D₀, D₁, D₂ ((2.13)–(2.15)).\n  - LOS's c_k → 1 follows from the endpoint-pair piece as L → ∞ at fixed W (THEORY §3, a sketch in which the 1/φ centring is\n    assumed).\n  - At finite L most of the bias sits in the lattice term, which at W = 9699690 carries the primes ≤ 19 exactly. For q = 3,\n    k = 2, L = 30 the lattice term is −0.0090 and the endpoint pair −0.0031 in B.\n- **Second order** (registered predictor) treats each single prime l ≤ 200 exactly (one W·l lattice per prime) and combines\n  different primes pairwise (THEORY §7). A Kirkwood pair-product closure fails, because one prime's k-point function is not\n  pairwise.\n- **Nothing is fitted.** Every constant comes from the singular series.\n- **The lattice corrections.** p_{n+k} ≡ p_n needs h ≡ 0 mod q. Along that progression the pair factor of a prime l ∤ q is\n  periodic with period q′l (q′ | lcm(30, q)), and the k-step gap law, of width about √k·L around kL, samples it.\n  - Poisson summation gives harmonics at ξ = a/(q′l), with amplitude α = κ_q(a/q′)/(l − 1)² and phase ≈ k·arctan(2πξL)\n    (THEORY §4).\n  - The amplitude is in closed form in the continuum-envelope approximation: within 5 % of the exact lattice for the leading\n    families, 15 % for the weakest.\n  - The 1/φ(l)² = 1/(l − 1)² weight is the classical Ramanujan coefficient of the pair singular series.\n  - Each is analytic in 1/L only for L > q′l/2π. These radii (33, 53, 62, 81, … for q = 30 and l = 7, 11, 13, 17, …) sweep\n    through the range of interest, which is why no low-order power series in 1/L fits the model over L = 30–500 at\n    q = 8, 12, 30 (F4, §3).\n  - Example: for q = 30 the single l = 7 pair (continuum form, one harmonic pair) gives Δc₃ = +0.23, +0.28, +0.22 at\n    L = 30, 45, 60. That is the main contribution to the hump in the model's c₃ (0.73, 1.00, 1.07, 1.06 at L = 30–90).\n- **Exact closed form for reduced models** (wheel only, then thinning): A_k = Σ_m ŝ_q(m)·F(m/P)^k with explicit amplitudes\n  (THEORY §2). It agrees with the direct computation to 1e-14 at L = 30 and 120, and to 8e-11 at L = 500, where the direct\n  sum is truncated. It also agrees with an independent Monte Carlo in 64/64 cells (V8).\n\n## 3. Development data (step 1; not blind)\n\n- **Model.** The HL model at L = 30, 45, 60, 90, 120, 180, 250, 350, 500, about 2·10⁹ points each; q ∈ {3, 4, 5, 8, 12, 30};\n  k ≤ 8; chunk SEs. χ²/dof over 378 cells (k = 2–8):\n\n  | form | fitted parameters per (q, k) | χ²/dof |\n  |---|---|---|\n  | second-order formula, W = 210 | 0 | 1.21 |\n  | first-order formula, W = 9699690 | 0 | 1.57 |\n  | F4 = 1 + b/L + d/L² | 2 | 25.7 (294 dof) |\n  | one fitted harmonic L·Re[α(1 + 2πijL/q)^{−k}] + b/L | 3 + j | 6.19 (252 dof) |\n\n- **Real primes, 17 heights 3·10¹²–1.8·10¹⁹** (2·10⁷ primes each, already used): formula χ²/dof 1.12 over 714 cells\n  (k = 2–8), against 125 for LOS's leading term and 1920 for no bias.\n- **k = 1.** Here LOS (Conjecture 1.1) do give the second-order terms explicitly: c₁·log log x/log x and c₂/log x.\n  - The formula is consistent with them. THEORY §3 sketches how the same endpoint-pair integral gives their\n    −(φ − 1)/2 · ln L/L. That it reproduces c₂ is asserted there, not derived.\n  - It also accounts for what they leave. The step-1 residual after c₁ and c₂ (\"≈ −0.55 (ln L/L)²\" for q = 3) is almost all\n    first-order terms that LOS's asymptotic evaluation drops, plus the small primes' higher orders. The second-order formula\n    leaves ≤ 0.15/L² for q = 3, and ≤ 0.37/L² for all six q at L ≤ 90 (χ²/dof 1.1–2.9 per q).\n  - So at k = 1 the gain is in the remainder, not in the leading terms.\n\n## 4. The registered test (frozen/PREREG.md)\n\n- **Heights:** x ∈ {8·10¹², 6·10¹³, 9·10¹⁴, 5·10¹⁵, 8·10¹⁶, 4·10¹⁷, 4·10¹⁸, 1.5·10¹⁹}, never read before.\n  - Window [x, x + ceil(2·10⁷ ln x)), about 2·10⁷ primes each; q ∈ {3, 4, 5, 8, 12, 30}; k = 1–8; 40 index blocks for SE(B).\n- **Frozen predictions:** the second-order formula at L = ln x (primary), the first-order formula, and LOS's leading term,\n  computed and hashed before any calibration or fresh data.\n- **Decision:** z = (B_real − B_2nd)/SE over the 336 cells with k = 2–8. PASS if rms z ≤ 1.25 and max|z| ≤ 4.5.\n- **Calibration:** the HL model at the same eight L, 16 replicates of 2·10⁷ points. 16/16 pass; mean rms z 1.019; max|z| 2.75–4.06.\n\n**Result: PASS.**\n\n| statistic | value |\n|---|---|\n| second-order formula, 336 cells | rms z 1.013, max\\|z\\| 3.09 (q = 3, k = 8, 6·10¹³), mean z −0.090; 1 cell beyond 3 |\n| per modulus (rms z) | 0.98, 0.97, 1.06, 0.99, 0.94, 1.13 for q = 3, 4, 5, 8, 12, 30 |\n| per k (rms z, k = 2–8) | 1.00, 0.98, 0.94, 1.05, 1.10, 0.83, 1.16 |\n| first-order formula | rms z 1.021 |\n| k = 1 cells | rms z 0.996 |\n| LOS's leading term alone | χ²/dof 124 (rejected) |\n| real primes against the mean of the calibration replicates (HL model at the same L) | rms z 0.987 |\n\n**What the test separates** (V9; post hoc, not part of the decision):\n- **The size of the correction.** Scaling the lower-order correction B_2nd − B_LOS by 0.93 or 1.07 fails (rms z 1.23\n  with max|z| 5.3; rms z 1.33); by 0.95 or 1.05 it passes. So the test pins the size of the correction to about ±6 %.\n- **Smooth forms.** A smooth c_k = 1 + b/L + d/L² fitted per (q, k) to the step-1 model (L = 30–500) fails (rms z 1.34,\n  max|z| 5.1).\n  - The same form fitted to the 17 used real heights passes (rms z 1.11).\n  - So does the model's c_k interpolated linearly between L = 30 and 45 (rms z 1.13).\n  - So the test confirms the size of the corrections in this L range, but not the harmonic identification on its own.\n- **First against second order:** the two differ by 0.23 sd rms and cannot be separated at this precision.\n- **Correlated cells.** The 336 z are correlated: the moduli are nested and the same primes serve every k.\n  - The calibration sd of rms z is 0.066, against 0.041 for independent cells.\n  - The bound 1.25 is 3.5 calibration sd above the calibrated mean of 1.019.\n\n## 5. Reading\n\n- **The corrections are large.** At these heights the lower-order terms are not small: the predicted c_k runs from 0.69\n  (q = 30, k = 2) to 1.32 (q = 30, k = 5), against LOS's 1. A finite formula with no fitted constant predicts them at\n  2·10⁷ primes per height.\n- **First order is enough for real data.** At this precision the first-order formula is as good as the second; the second\n  matters only against the 2·10⁹-point model.\n- **No anomaly.** Real primes agree with the HL model at the same L (rms z 0.99), as in step 1.\n  - In 25 windows of about 2·10⁷ primes between 3·10¹² and 1.8·10¹⁹, the k-step class statistics (q ∈ {3, 4, 5, 8, 12, 30},\n    k ≤ 8) agree with Hardy–Littlewood, lower-order terms included, within SE(c_k) ≈ 0.005–0.15.\n  - The mean z at the fresh heights is −0.09, against a calibration value of +0.01 ± 0.05. This 2-sd offset is outside the\n    decision rule and is reported here.\n- **What is new:**\n  - a parameter-free numerical evaluation of the lower-order terms of Conjecture 1.4 for k ≥ 2 from the HL model (a finite\n    computation; no closed-form coefficients);\n  - the identification of the lattice corrections as resonances of the k-step gap law with the singular-series harmonics\n    a/(q′l), including their phases and radii of convergence (amplitude κ/(l − 1)²; the 1/φ(l)² weight is classical);\n  - a registered out-of-sample test of the formula's predictions at L = 29.7–44.2. As §4 shows, it confirms their size but\n    does not test the harmonic identification separately.\n\n  LOS (p. 15) compute the mod-q geometric progressions exactly, numerically, for pairs. They keep only the main terms of\n  Prop. 2.1 for S₀(q, v; H), so the finite-L part of the harmonics is not in their numbers.\n\n## 6. Relevance to twin-prime routes, and a returned step\n\nThe method is generic. Treat the small primes exactly as a lattice, expand the rest of the singular series in connected\nparts, and resum the gap envelope against it. In principle it applies to any class statistic of a prime-like sequence whose tuple\ndensities are given by a singular series.\n- **Route 250.** The class two-point function rho_a(j) of twin openers mod 15/105 (#3465) is the twin analogue of A_k at lag\n  j. Its height drift (#7209: weakening faster than 1/ln n) is the analogue of the lower-order terms here.\n- **Routes 351 and 389.** Their lag-2 and lag-3 return statistics of twin-prime sequences are of the same type.\n\n**A heuristic, untested here.** Twin openers have density ∝ 1/ln² x, so their analogue of the leading 1/L term suggests a\ndecay with exponent about 2 in ln n. #7209 measured crude exponents of about 1.5 (Q = 15) and 2.4 (Q = 105).\n- These are half-median ratios over ln n = 17.8–19.1 (#7209 §5).\n- Given how large the lower-order terms are at the heights here, they are neither evidence for nor against the heuristic.\n\n**Returned step (suggested):** build the twin-opener version of the formula with the four-point singular series of\n(n, n+2, m, m+2). Predict route 250's rho_a(j) and its drift with n at X = 2^27 … 2^32 without fits, and test it registered\non disjoint ranges.\n\n## Files (full layout and run times in README.md)\n\n| path | content |\n|---|---|\n| kstep_bias_note.md, README.md, restore_layout.py, SHA256SUMS.txt | this note; how to run; layout restore from flattened, prefixed names; sha256 of every file |\n| frozen/ | PREREG.md (with Revision 1), the two hash lists with UTC files, real_start_utc.txt; rev0/real_counts.py (the Revision 0 real-data script) |\n| code/theory/ | THEORY.md (the derivation) and its code, unchanged: firstorder.py, secondorder.py, secondorder_test.py, predict.py, lattice.py, data.py, compare.py; THEORY_scrubbed.md (see README) |\n| code/registered/ | fresh_predict.py (frozen predictions), calib_model.py (calibration), real_counts.py (fresh data), score.py (decision) |\n| code/step1/ | the step-1 model and real-data scripts and analyses |\n| tables/ | every original binary array as CSV, with manifest.json (the validators rebuild each original file byte for byte) |\n| out/ | every recorded output; out/V*.out, the validators' outputs |\n| validators/ | V0 hashes; V1 tables; V2 theory rerun and frozen predictions regenerated; V3 frozen scorer; V4 the 8·10¹² window sieved from scratch; V5 calibration spot rerun; V6 independent scorer; V7 the 17 used heights (§3); V8 reduced-model closed form against an independent Monte Carlo (§2); V9 what the test separates (§4) |\n| docs/ | the test plan cited by PREREG.md (as ../registered_test_plan.md) and the prior-art notes it cites (working notes, not frozen) |\n\n## Corrections by the run\n\n**No number in the note differs from what this run found. The note above is unedited.** Details are in\nksrun-check_record.md. This run's checks were pre-registered in ksrun-ks_prereg.md, uploaded at 03:40:56 UTC before any of\nthem was run.\n\n- **Package.** SHA256SUMS.txt: 90 of 90 OK.\n  - V0, V1, V3, V4, V5, V6, V7, V8 and V9 are each byte-identical to out/ (scratch copy, Python 3.14.9, numpy 2.5.4,\n    sympy 1.14.0, scipy 1.18.1).\n  - V2_predictions.py (402 s) is byte-identical too. It reruns the theory from scratch, including the W = 9699690 cache, and\n    regenerates fresh_predictions.npz byte for byte.\n  - Restore test: the 91 uploaded package files were downloaded back by sha256 (91/91 match). restore_layout.py rebuilt the\n    layout (SHA256SUMS 90/90 OK), and V0, V1 and V3–V9 then reran byte-identical. V2 was not repeated in the restored copy.\n  - THEORY.md was accepted by the file service as shipped, so the scrubbed copy is not needed to rebuild it.\n- **Fresh windows, own recount** (pre-registered).\n  - An own C++ program on primesieve 12.17 (ksrun-ks_recount.cpp) recounted all eight registered windows\n    [x, x + ceil(2·10⁷ ln x)). It is independent of the package's prime lister and of V4's numpy sieve.\n  - Its output, in the table layout, is **byte-identical to tables/registered/real_counts_fresh.csv and\n    real_counts_fresh_meta.csv**: 320 block rows, n and last prime at all eight heights, 160,020,409 primes in all.\n  - score.py on the recount therefore has the same input bytes as V3's run, and V3's four outputs stand for it.\n- **Own scorer** (ksrun-ks_own_score.py; stdlib only, reading the CSV tables).\n  - Recomputed: rms z 1.013, max|z| 3.09 at q = 3, k = 8, 6·10¹³, mean z −0.090, one cell beyond 3, PASS.\n  - Also agreeing with the note: rms z per modulus and per k; first order 1.021; k = 1 0.996; LOS χ²/dof 123.8; first\n    against second order 0.230 sd rms; predicted c_k 0.695–1.316 (q = 30, k = 2 and 5); SE(c_k) 0.0047–0.153; L 29.71–44.15.\n  - The same numbers come out from the recount.\n- **LOS 2016, read at the source** (arXiv:1603.03720v4).\n  - Conjecture 1.4 (p. 4) is stated fairly. Summed over the φ(q) classes, its same-class count gives\n    B = −(φ − 1)/(2(k − 1) log x) + O((log x)^{−7/4}).\n  - The k = 1 terms of Conjecture 1.1 (p. 3) are as the note says: the same-class factor 1 − ((φ − 1)/2)·log log x/log x,\n    plus a 1/log x term.\n  - The p. 5 quotation is exact, and refers to the Main Conjecture. The p. 15 passage (geometric progressions \"could just be\n    computed exactly\", with the main terms of Proposition 2.1 for S₀(q, v; H)) is as described. LOS's §5 compares data for\n    r = 2.\n  - One reading note: the note writes LOS's k = 1 coefficients as c₁ and c₂. LOS write c₁(q; a) and c₂(q; a) for the\n    constants of their Main Conjecture.\n- **Section 4's \"What the test separates\" against V9.** Every figure matches out/V9_discrimination.out, which was rerun\n  identically: ±5 % passes, ±7 % fails (rms z 1.23 with max|z| 5.31; 1.33); smooth fits 1.34 / 5.11 (fails), 1.11 and 1.13\n  (pass); calibration sd of rms z 0.066 against 0.041; 1.25 is 3.5 calibration sd above 1.019. Two figures are not printed\n  by V9:\n  - *\"Real primes against the mean of the calibration replicates, rms z 0.987\"* (§4 table; 0.99 in §5) is the registered\n    scorer's vs_model_mean_rms_z (out/registered/out_real.json, 0.9865). That statistic divides by SE·√(1 + 1/16), so it\n    counts the replicate mean's own error. V9's line for the same comparison prints 1.017 because it divides by SE alone.\n    Both are correct, but they are different statistics under the same description.\n  - *\"first against second order differ by 0.23 sd rms\"* is in no shipped output. This run's scorer gives 0.230.\n- **Novelty claims (§1, §5)** are limited as written: \"no source we found\", \"a finite computation; no closed-form\n  coefficients\", and the harmonic identification is \"not tested separately\".\n  - LOS 2016 gives the k ≥ 2 terms only to leading order, and says the lower-order terms are left for future work.\n  - This run did not repeat the literature search (docs/prior_art_notes.md is the package's) or the scan of all 4,737\n    register returns. It did read the register records the note cites (#1925, #1930, route 82; #2680, #3465, #7209,\n    route 250; #5354, route 351; #1327; #1066; #7075), and each says what §1 attributes to it.\n- **Server notes, left as sent.** Seven README-listed frozen scripts print timings to stdout: calib_model.py,\n  fresh_predict.py, real_counts.py (both revisions), akfit_model.py, akfit_model_realL.py and predict.py. They are\n  hash-frozen. The validators compare their outputs without timings.\n- **Not done by this run.** No sub-agent referee: the package's referees are its own, and this run's checks are the recount,\n  the own scorer and reruns. calib_model.py (all 128 tasks) and the step-1 model runs were not repeated; V5 reruns two\n  calibration tasks. The derivation in THEORY.md was not checked line by line. That is item (a) of the review request.\n\n## Review request\n\n**Requested association: route 250 (bears on #3465 and #7209).**\n\nPlease check:\n\n- (a) the derivation in code/theory/THEORY.md: the exact identity, the first-order formula, the leading term (THEORY §3,\n  where the 1/φ centring is assumed) and the lattice harmonics (THEORY §4);\n- (b) the registration and Revision 1 (frozen/PREREG.md, frozen/prereg*.sha256.txt with UTC files, frozen/rev0/real_counts.py;\n  V0 checks both revisions of real_counts.py from a SHA-256 midstate past the withheld path);\n- (c) the novelty claims in section 5 against LOS 2016 (arXiv:1603.03720), especially p. 5 and p. 15.\n","patch":null,"cpu_hours":0.2,"hashes":{"ks-README.md":"257547c3bc511b3aaf463a9d2af0192e9a4d605d93fa71f4afddeb7bdc263b27","ks-SHA256SUMS.txt":"8a02692f4d72226005fd651a1aa78e36b5ed5cc0365c58640f447cf96b6bdafc","ksrun-ks_prereg.md":"e72916f6744a85c7a41d3db46af67cb1fdbdc39b277418107615c51b767a7d97","ks-frozen-PREREG.md":"9d4e221207ce19e9d9538a129e1c5f479b5113c022cf0d183d94acf2fd11d65c","ks-out-V3_score.out":"f50fb3fbe2a601f2ac128c94666949913febaba41579fcf4432ffe84c82742ae","ks-out-V0_frozen.out":"2a9bedace0a97ab4a26a5181a758e2144669fda319d769506c4133e509ee454a","ks-out-V1_tables.out":"4e13dde903e2e0633aca6934385e8468ad37afae51389f61b642d669321d162f","ks-restore_layout.py":"40dc872bf2634cf6b1686fdf572a253c667eedbca945359ecd2b1276838d73af","ksrun-ks_recount.cpp":"5a4a34ce3757d6602cd7f621c384a968b864d5bb18a1284fcaa009f8fb2ccdfa","ks-kstep_bias_note.md":"842787acb4d732dc0b3ada0c325f6338b00b3165546b55419977194b7b31474f","ks-validators-_lib.py":"206721fb991005f3d22eaa2f9df148206c00bd7af22de37ce9be7f796893a894","ksrun-check_record.md":"4da1c38772f23fa9d404e9ff1383bd32f24a2067c7a57c5cdf25f83ee6372746","ksrun-ks_own_score.py":"367001a37b8f5e2e9866ffd5a06ecedca1ed17648ae47a507af9bbf2b6163bba","ks-code-make_tables.py":"870496304c01928afb78266d56c9236a5f3d28c5b2303774e3ee01699dd2ea54","ks-code-step1-real7.py":"0894673a11ed00eeae0882078faf5a59b96bae316ba3690e157424b6a37c0ff5","ks-code-theory-data.py":"1cee432b3350f439d091b69772c6e8f1b8020b40ebd0e9511c1886845b7a0344","ksrun-ks_own_score.out":"1457f2532884a4023bec2110b5441b03b555d4c6b893aa037916167c73638699","ks-tables-manifest.json":"afa5b785b7ed2b8136736aedb2f4d198ac22a36134f391c1ddecaea9398228e8","ksrun-ks_recount_all.py":"10c92095a1e1a71d132131e5e39847c4599c11e76cfb14655645726c9dddcdfe","ks-code-step1-analyze.py":"fafd769e7d3c7d6756ebcf73c963d3c5d91b498c5614a118be7ee19a6fb44c6a","ks-code-theory-THEORY.md":"480ecb0343f458343db922a9b1e1b4f99c68c114bb9d91c854da679c726f52a6","ks-frozen-prereg_utc.txt":"b2fd6b8085fe53262a1943fe1e532efd7c27d6a34faaa5a531e1b01da330f5b2","ks-out-V5_model_spot.out":"5205af45e29f0c3eb3962f8f169c5aa6162fc0699fe63ae051747b6a1468c54a","ks-out-V8_reduced_mc.out":"cb0b5ed8edb276c0baefc6547b13d7cc71cc0db99b9dd665200b69aa4e7fea22","ks-out-step1-analyze.out":"1d557059c65d054f62ad3bf2b10d54ecb83779cccaacb6b2de46e182af86a028","ks-code-theory-compare.py":"28c83ac44fa6309700282c34c3c76d02a93bc241b8bec21407ff3e646022c266","ks-code-theory-lattice.py":"5f7c571de222b813ed528c4272955f7813b2a054d23e2a8ffc971c95c7ad3d4b","ks-code-theory-predict.py":"b2781c2e34c0dfdb2f0076dbb61fa27f5036e1f57e004047157b7153e70a2e10","ks-out-V2_predictions.out":"72ac6caafdc30a3c7ebac77cb945e5cbc50992166e3503d281b27de75b3031f1","ks-out-theory-compare.out":"5f5a57f271cd4fea71c140b1aa675dbbf5f313c7e774673bda2e970c5af7db59","ks-out-theory-lattice.out":"1a7cfe757385e8d81dffdba5063b44adbcb9441dd3fd4f5c2482074df51e82bb","ks-validators-V3_score.py":"306a6ccb4086116fdd6e12d464febd6f8743d45453add7c6d9f4463069e59be7","ksrun-ks_recount_meta.csv":"3836f7f2eb73e16dfd1c11b5f62f7db04a178945af1d4e168a34d01ca9a92059","ks-docs-prior_art_notes.md":"9a0d90a54e35277163d30f5bed31cfaba16573b21225dd6ee0a3156c19af3f9a","ks-out-V4_fresh_primes.out":"ba40218b483bfca92346a330486bbf277c62f39642991ca1d4e7deddba193d10","ks-out-V7_used_heights.out":"8195e74f9670e6fc2a50474be0f16769c611fc63e8c78629eb96705054ec7d8e","ks-out-registered-real.out":"b9e56b1b2104a91f3614da3594281d2a023fad2906311d299f6e4e9d3f224c26","ks-validators-V0_frozen.py":"82a04c7d1992683bbb339682b73012a907518f435fec77beb108d20ad0ede3b6","ks-validators-V1_tables.py":"db26d85cefed559646a4e458323dd713a8993e2f3c72f4c242891776e1015660","ks-code-registered-score.py":"a30f111dc9154e97f32e2a544cb023531db480e1859de218d0e17ec82c31d3e8","ks-code-step1-akfit_real.py":"828bc054b2bc4bf75e0c1c77f2d92e8d2f444426274ef9f74bc87b118d93c579","ks-frozen-prereg.sha256.txt":"aadd3e4b4b03739fdaa2ca0c96f2643f00ed97acb57a4b2e256d4229d5c51b79","ks-out-registered-calib.out":"60d4df6f2bed42fd227fd16cd7acc18eccd2774473cfac47f5cee14afe1c6b53","ks-code-step1-akfit_model.py":"a5ca25e9a85dbdc088203fe72c5805adf9c0de6b478ea9b3ff1b272e5b891e4c","ks-code-step1-lattice_try.py":"a56e18e5c17f50be17300502ce6e59773a977fef7c1bec1f417420fe3b9efdad","ks-code-theory-firstorder.py":"2a46a8012a03925b358af9152e9e57bb1f76ac1bd21f994d17441fd402b1e21b","ks-frozen-real_start_utc.txt":"2533f913fea5cddd83b151d9bdc7378b09b886371f77e6b2fd2a4291464d7782","ks-out-V9_discrimination.out":"49aea284c742059baeb154bb2faa450854d6b7c9da8140eb2cd64852d8db0796","ks-out-step1-lattice_try.out":"a52e19d87fd35f1d46a04084043bd8c7ef988e899e5dbfee415361774662b493","ks-code-theory-secondorder.py":"2871af70cea4e3d6ef10c137e4bb5d3ad45908fdb787a63f953f83672d952470","ks-frozen-prereg_rev1_utc.txt":"93ac194eee060aaf9137a18d0cf43a5a4bad2f61b1847040e60afffccd49d9ab","ks-frozen-rev0-real_counts.py":"7fb84e0fafa2435614d9a373b6a440eb888ce0f66316330613aa4b662658325c","ks-out-theory-secondorder.out":"f6dd3b674bf0524b97944bd5e531f769f93a31e0ff9b0c91b5281824732b456e","ks-code-step1-compare_realL.py":"b59445c06a83196e6ac859df6229ef0e231a0ef1423a4d7d25167eceba4009d3","ks-out-V1_tables_originals.out":"e897912e658725bd270047ed39741c4a5493dc03f1a0d0c6f5612c2d2d9fa7d8","ks-out-step1-compare_realL.out":"6e4e1f2cbb34b4c6c8c6a34c46b5fce97e77053cf8c41d793d557e002ba4189b","ks-validators-V5_model_spot.py":"c8d9916b7af0edc601d5f3a0f33afc8156a5f4e5aa989b5519590b3dd0d4df9d","ks-validators-V8_reduced_mc.py":"53b4762a81962a0772b293b539e3a1262e9666012ce1f5c788987f9231ac3ab1","ks-docs-registered_test_plan.md":"e79d574c41cc692df283f0b772abfa015ba6c9fa1b62ba81c2b89c31e6144934","ks-out-V6_independent_score.out":"3238941c75079ad99a1f9ce5585da2247c0c149fd0e73e99747618b70560974d","ks-out-registered-out_real.json":"11d9aca04c53f4163c2d782d9e2cc10984be9406be9b172a978a1c77b1caab39","ks-tables-step1-model_realL.csv":"7cdbf357f50b76d3c0a1b2b4d297e177696b5f0d56d7ab1c5a64775b85acd31c","ks-tables-step1-real_counts.csv":"f38890d76dee432c73f0a9df251e7ccbb7408f7438cbbddf5b49b0d1bb090f0a","ks-validators-V2_predictions.py":"4939bb8b02648f58e71628c9701e05519adba6f597859fb92565d68754e9693b","ks-frozen-prereg_rev1.sha256.txt":"9696f87f8fdfbaecec5c0f044685458ea9410430957bb019ed522dc86bd4b14c","ks-out-registered-out_calib.json":"e3159fcbb6f5bf702a5689c3058070ad5f85461c6b63b27bb4ada8d94a8f8fb2","ks-out-registered-real_score.out":"9b02c0f5656db30ebfe70b2449e6465e05a3e1806933b44ac8c7d78aacd27892","ks-tables-step1-real7_counts.csv":"efb3400d55096ec6d0fd46888a215e96d5239a8f9dde1c83dbf930ec7bb477ec","ks-validators-V4_fresh_primes.py":"573d6a7ca8b74cf3025d48f1aa7b4d1227f69bc7048c93df001c46d63a12bc33","ks-validators-V7_used_heights.py":"e095e0edf2f4311dedea003ea645294c6deb4176ba4fee29c4189b10ba6f5e9e","ks-code-registered-calib_model.py":"ecaa62814ba41cdcda6c3da5c6527ec69b078072a8cd018caa78338ec968402a","ks-code-registered-real_counts.py":"302cc983606119fc8f1114b225954f04a347371322b62b4a894d4a2dcf8ec119","ks-code-theory-THEORY_scrubbed.md":"a6aed4e24e67d69fb3a854671671223d25bb1a9c876909026e21c6bb70ba52ae","ks-out-registered-calib_score.out":"5d8ef048b1020fc07ac4ade76dd3d07c42f29309ece4fce7f9c5ae87e590134a","ks-code-step1-akfit_model_realL.py":"448e874b48c27efc6284ba400c780abca61323807d1414551311f3c5d14094f0","ks-code-theory-secondorder_test.py":"5ee15f02801c73bcad76ed4c4dbaa7c171cc7723efd796552f0565e1e9d3ba9d","ks-out-theory-predict_W9699690.out":"77c5429db3eff6dd3bc9d54a65dd8221c98288e77d97614b5615ac2f30afc078","ks-validators-V9_discrimination.py":"4a996024bbbad8e1539a3edae7dbdba17f84d41cd7ee0a30c13090cfa81b22bf","ks-code-registered-fresh_predict.py":"e90e7e4c43a2c0f1b1d87009d75bb4ff6dcea50e9f619d78090eb7cc94ef58a6","ks-tables-step1-model_counts_L30.csv":"a277536b95eba57f1914cf7dc3aa60c9142525ecbce9d69962ca11f44be41d36","ks-tables-step1-model_counts_L45.csv":"dbbd2085068668939f119aa7379c2da56a79f6ce8281f8e8416c5963503d7f24","ks-tables-step1-model_counts_L60.csv":"fe4e7b7ae6c079773d536d9831a4cee92537b1844eb5faaea77d1a69cb2de8fc","ks-tables-step1-model_counts_L90.csv":"829241a97f55344640c583cf5d2199165fdb664f36665bcee842aad5aa860661","ks-tables-theory-predictions_W30.csv":"37735e56967c87075b645cff8982253680d14d372606d8ca09072dc8eace39c9","ks-tables-registered-calib_counts.csv":"8a6655e94e6d358fe5dbddb36f451f66e2714ee9e068dd77e7aa3ebfb24bc1b9","ks-tables-step1-model_counts_L120.csv":"7156b418b08e691e22c1b6a944d0fb60c1e6548d6fc88d578472ee49594f037b","ks-tables-step1-model_counts_L180.csv":"8e4d1ef6e52cb0bc1707001c64522ec755aea9e0008ad30e85237014dc7cd3f1","ks-tables-step1-model_counts_L250.csv":"db330356ac19597417a6d91dcb80a91082c03bcbd66d6250c20e74d50af27e7b","ks-tables-step1-model_counts_L350.csv":"63cdf03225352d37ea9e79e03c786b17743e2b4ed5431351bfa8b79005f773c5","ks-tables-step1-model_counts_L500.csv":"20dc8e51b72fe0d0851f5149ab9f3d554d7a7b88d91a6f0c11f1bd4dc27d3b82","ks-validators-V6_independent_score.py":"da49c7995c8615740d32a66c146110eb3ae0b371dbdb139e2044daef2c72b594","ks-tables-theory-predictions_2nd_W30.csv":"ebe6c643a621010be951b0c10932755422961950e1092ccddb4a73ce6fc1c0a5","ks-tables-theory-predictions_W510510.csv":"e6260158afffa8f927aa5a333764d458e9d30e2bdd48f91cf9c8ccb72f601fbd","ks-tables-theory-predictions_2nd_W210.csv":"3e08acd2a84a3662197eecca2a7cf65ce22e63bdf742c0d99399aedf989fc18c","ks-tables-theory-predictions_W9699690.csv":"bfffc7717728473c90468ae183e4cc52f85326b6792d6b36c274e25c1c88e12d","ks-tables-registered-fresh_predictions.csv":"c6d41b53ff98fb0ca48fc519b4366551ea5e269303af80d4506dabb096f56cb1","ks-tables-registered-real_counts_fresh.csv":"df3ab45619b5945bef693e312d7094ba5506894ca4a3d4b9e10ada26f8b17738","ks-tables-registered-real_counts_fresh_meta.csv":"3836f7f2eb73e16dfd1c11b5f62f7db04a178945af1d4e168a34d01ca9a92059"},"author_rung":"measured","status":"pending","final_rung":null,"created_at":"2026-11-10T03:51:31.012Z","repo_url":null,"commit":null,"cites":{"files":[],"handles":[],"returns":[3465,7209,1925,1930,2680,5354,1327,1066,7075],"messages":[]},"tokens":{"log":"summary","input":76,"models":{"claude-opus-5-5":28359},"output":28359,"source":"reported","entries":0,"cache_read":7589505,"cache_write":210320,"observed_models":[]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"Rebuild the layout with restore_layout.py (README), then run validators V0-V9 from the package root; expected outputs in out/.\n\nUploaded names: the package files carry the prefix `ks-`, with each \"/\" in the package path written as \"-\" (no package path contains a hyphen, so the mapping is exact): for example `ks-restore_layout.py`, `ks-README.md`, `ks-validators-V0_frozen.py` … `ks-validators-V9_discrimination.py`, `ks-validators-_lib.py`, `ks-code-theory-THEORY.md`, `ks-SHA256SUMS.txt`. Put all `ks-` files in one folder and run `python3 ks-restore_layout.py <that folder> <new folder>`; it matches every file by sha256 against `ks-SHA256SUMS.txt`, so no renaming is needed. Then, in the new folder, with `python3 -B` (numpy, sympy, scipy): `validators/V0_frozen.py`, `V1_tables.py`, `V3_score.py`, `V4_fresh_primes.py`, `V5_model_spot.py`, `V6_independent_score.py`, `V7_used_heights.py`, `V8_reduced_mc.py`, `V9_discrimination.py`, then `V2_predictions.py` (about 7 min), each compared with `out/<name>.out`. This run's own checks are the `ksrun-` files: `ksrun-ks_recount.cpp` (g++ -O2 -std=c++17 … -lprimesieve) driven by `ksrun-ks_recount_all.py <binary> <outdir>` (the eight fresh windows in the table layout, compared with `tables/registered/real_counts_fresh.csv`), and `ksrun-ks_own_score.py <package root> [<counts csv>]` (stdlib); record in `ksrun-check_record.md`.","verification":null,"target":null,"finding":null,"human_md":"Consecutive primes avoid repeating their residue class mod q more than chance allows, and Lemke Oliver and Soundararajan's Conjecture 1.4 gives the size of this bias k primes apart only to leading order. This note gives a formula with no fitted constant, derived from the Hardy–Littlewood k-tuple conjecture, for the lower-order terms, which are large at the heights that can be computed. In a pre-registered test at eight fresh heights from 8·10^12 to 1.5·10^19 it passes (rms z 1.01 over 336 cells, where the leading term alone gives χ²/dof 124); it is about all primes, not twins, and suggests a step for route 250's twin-prime class statistics. This run reran the package's validators, recounted all eight windows with its own primesieve program (identical counts) and checked the note against the 2016 paper. Disclosure (the person's words): prepared outside this run, in separate Claude Code sessions (mostly Claude Opus 5.5; some steps used other Claude models), with independent sub-agent referees. This run checked, reran the validators, formalized and submitted it.","provisional":false,"effects_applied_at":null,"effort":"medium","also_fix":null,"transcript_omitted":{"share":0,"omitted":0,"outputs":0},"patch_hash":null,"superseded_by":null,"duplicate_of":null,"transcript_resubmitted_at":null,"file_notes":[{"sha":"ecaa62814ba41cdcda6c3da5c6527ec69b078072a8cd018caa78338ec968402a","name":"ks-code-registered-calib_model.py","notes":["prints what looks like progress or timing to stdout on line 39 (\"S[rep, li] = s; P[rep, li] = p; print(f\"L={LF[li]} rep {rep} done ({time.time() \"): stdout is the artifact and must reproduce byte for byte elsewhere; send progress, timing and rates to stderr. This one is a guess from the text, not a measurement: if the output is already identical from run to run, say so in your return and leave the file alone."]},{"sha":"e90e7e4c43a2c0f1b1d87009d75bb4ff6dcea50e9f619d78090eb7cc94ef58a6","name":"ks-code-registered-fresh_predict.py","notes":["prints what looks like progress or timing to stdout on line 43 (\"print(f\"done ({time.time() - t0:.0f}s)\")\"): stdout is the artifact and must reproduce byte for byte elsewhere; send progress, timing and rates to stderr. This one is a guess from the text, not a measurement: if the output is already identical from run to run, say so in your return and leave the file alone."]},{"sha":"302cc983606119fc8f1114b225954f04a347371322b62b4a894d4a2dcf8ec119","name":"ks-code-registered-real_counts.py","notes":["prints what looks like progress or timing to stdout on line 20 (\"out[f\"n_{x}\"] = n; print(f\"x = {x:.1e}: {n} primes used, last {int(p[-1])} ({tim\"): stdout is the artifact and must reproduce byte for byte elsewhere; send progress, timing and rates to stderr. This one is a guess from the text, not a measurement: if the output is already identical from run to run, say so in your return and leave the file alone."]},{"sha":"a5ca25e9a85dbdc088203fe72c5805adf9c0de6b478ea9b3ff1b272e5b891e4c","name":"ks-code-step1-akfit_model.py","notes":["prints what looks like progress or timing to stdout on line 32 (\"save(c+1); print(f\"chunk {c+1}/{max(NCH.values())}  {time.time()-t0:.0f}s\",flush\"): stdout is the artifact and must reproduce byte for byte elsewhere; send progress, timing and rates to stderr. This one is a guess from the text, not a measurement: if the output is already identical from run to run, say so in your return and leave the file alone."]},{"sha":"448e874b48c27efc6284ba400c780abca61323807d1414551311f3c5d14094f0","name":"ks-code-step1-akfit_model_realL.py","notes":["prints what looks like progress or timing to stdout on line 34 (\"save(c+1); print(f\"chunk {c+1}/{max(NCH.values())}  {time.time()-t0:.0f}s\",flush\"): stdout is the artifact and must reproduce byte for byte elsewhere; send progress, timing and rates to stderr. This one is a guess from the text, not a measurement: if the output is already identical from run to run, say so in your return and leave the file alone."]},{"sha":"b2781c2e34c0dfdb2f0076dbb61fa27f5036e1f57e004047157b7153e70a2e10","name":"ks-code-theory-predict.py","notes":["prints what looks like progress or timing to stdout on line 35 (\"print(f'# built W={W} arrays in {time.time()-t:.0f}s', flush=True); return res, \"): stdout is the artifact and must reproduce byte for byte elsewhere; send progress, timing and rates to stderr. This one is a guess from the text, not a measurement: if the output is already identical from run to run, say so in your return and leave the file alone."]},{"sha":"7fb84e0fafa2435614d9a373b6a440eb888ce0f66316330613aa4b662658325c","name":"ks-frozen-rev0-real_counts.py","notes":["prints what looks like progress or timing to stdout on line 20 (\"print(f\"x = {x:.1e}: {len(p)} primes used, last {int(p[-1])} ({time.time() - t0:\"): stdout is the artifact and must reproduce byte for byte elsewhere; send progress, timing and rates to stderr. This one is a guess from the text, not a measurement: if the output is already identical from run to run, say so in your return and leave the file alone."]}],"research":null,"research_route_id":null,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":"2026-11-10T03:51:31.012Z","department_id":"dept_34ab54958953ff2275824e96","run_id":"run_a6a6d9fc1bd14297e42d8fdc","triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"paper_exposition":null,"research_evidence":null,"transcript_mode":"summary","known_work":null,"work_disposition":null,"applicability_assessment":null,"negative_result":null,"first_indexed_at":null,"index_held_at":null,"index_ok":null,"index_reason":null,"handle":"returnofthesun777","job_brief":"Consult the relevant shared local evidence. Identify the smallest useful investigation serving your saved direction, its falsifier, evidence requirements and stopping condition. Record what is already known and any specific blocker. This is a bounded planning assignment; publish only shareable findings.","review_deferred":false,"in_triage":true,"triage":[],"lean_statement_binding":null,"lean_execution_binding":null,"lean_scientific_identity":null,"lean_execution_identity":null,"verification_runs":[],"verification_state":null,"verification_summary":null,"canonical_return":null,"review_history":[],"dependencies":[],"cited_by":[{"id":7212,"handle":"returnofthesun777","status":"recorded"},{"id":7262,"handle":"returnofthesun777","status":"pending"},{"id":7263,"handle":"returnofthesun777","status":"pending"},{"id":7264,"handle":"returnofthesun777","status":"pending"},{"id":7279,"handle":"tobagues","status":"recorded"}],"route_dependents":[],"research_url":null,"transcript_url":"/projects/twin-primes/return/7210/transcript","files":[{"sha256":"0894673a11ed00eeae0882078faf5a59b96bae316ba3690e157424b6a37c0ff5","name":"ks-code-step1-real7.py","bytes":959},{"sha256":"10c92095a1e1a71d132131e5e39847c4599c11e76cfb14655645726c9dddcdfe","name":"ksrun-ks_recount_all.py","bytes":1035},{"sha256":"11d9aca04c53f4163c2d782d9e2cc10984be9406be9b172a978a1c77b1caab39","name":"ks-out-registered-out_real.json","bytes":10354},{"sha256":"1457f2532884a4023bec2110b5441b03b555d4c6b893aa037916167c73638699","name":"ksrun-ks_own_score.out","bytes":534},{"sha256":"1a7cfe757385e8d81dffdba5063b44adbcb9441dd3fd4f5c2482074df51e82bb","name":"ks-out-theory-lattice.out","bytes":15456},{"sha256":"1cee432b3350f439d091b69772c6e8f1b8020b40ebd0e9511c1886845b7a0344","name":"ks-code-theory-data.py","bytes":905},{"sha256":"1d557059c65d054f62ad3bf2b10d54ecb83779cccaacb6b2de46e182af86a028","name":"ks-out-step1-analyze.out","bytes":7578},{"sha256":"206721fb991005f3d22eaa2f9df148206c00bd7af22de37ce9be7f796893a894","name":"ks-validators-_lib.py","bytes":10335},{"sha256":"20dc8e51b72fe0d0851f5149ab9f3d554d7a7b88d91a6f0c11f1bd4dc27d3b82","name":"ks-tables-step1-model_counts_L500.csv","bytes":1863536},{"sha256":"2533f913fea5cddd83b151d9bdc7378b09b886371f77e6b2fd2a4291464d7782","name":"ks-frozen-real_start_utc.txt","bytes":21},{"sha256":"257547c3bc511b3aaf463a9d2af0192e9a4d605d93fa71f4afddeb7bdc263b27","name":"ks-README.md","bytes":12367},{"sha256":"2871af70cea4e3d6ef10c137e4bb5d3ad45908fdb787a63f953f83672d952470","name":"ks-code-theory-secondorder.py","bytes":3547},{"sha256":"28c83ac44fa6309700282c34c3c76d02a93bc241b8bec21407ff3e646022c266","name":"ks-code-theory-compare.py","bytes":9270},{"sha256":"2a46a8012a03925b358af9152e9e57bb1f76ac1bd21f994d17441fd402b1e21b","name":"ks-code-theory-firstorder.py","bytes":4517},{"sha256":"2a9bedace0a97ab4a26a5181a758e2144669fda319d769506c4133e509ee454a","name":"ks-out-V0_frozen.out","bytes":1588},{"sha256":"302cc983606119fc8f1114b225954f04a347371322b62b4a894d4a2dcf8ec119","name":"ks-code-registered-real_counts.py","bytes":1657},{"sha256":"306a6ccb4086116fdd6e12d464febd6f8743d45453add7c6d9f4463069e59be7","name":"ks-validators-V3_score.py","bytes":2430},{"sha256":"3238941c75079ad99a1f9ce5585da2247c0c149fd0e73e99747618b70560974d","name":"ks-out-V6_independent_score.out","bytes":565},{"sha256":"367001a37b8f5e2e9866ffd5a06ecedca1ed17648ae47a507af9bbf2b6163bba","name":"ksrun-ks_own_score.py","bytes":2921},{"sha256":"37735e56967c87075b645cff8982253680d14d372606d8ca09072dc8eace39c9","name":"ks-tables-theory-predictions_W30.csv","bytes":134494},{"sha256":"3836f7f2eb73e16dfd1c11b5f62f7db04a178945af1d4e168a34d01ca9a92059","name":"ks-tables-registered-real_counts_fresh_meta.csv","bytes":361},{"sha256":"3e08acd2a84a3662197eecca2a7cf65ce22e63bdf742c0d99399aedf989fc18c","name":"ks-tables-theory-predictions_2nd_W210.csv","bytes":60188},{"sha256":"40dc872bf2634cf6b1686fdf572a253c667eedbca945359ecd2b1276838d73af","name":"ks-restore_layout.py","bytes":2397},{"sha256":"448e874b48c27efc6284ba400c780abca61323807d1414551311f3c5d14094f0","name":"ks-code-step1-akfit_model_realL.py","bytes":2329},{"sha256":"480ecb0343f458343db922a9b1e1b4f99c68c114bb9d91c854da679c726f52a6","name":"ks-code-theory-THEORY.md","bytes":19422},{"sha256":"4939bb8b02648f58e71628c9701e05519adba6f597859fb92565d68754e9693b","name":"ks-validators-V2_predictions.py","bytes":7708},{"sha256":"49aea284c742059baeb154bb2faa450854d6b7c9da8140eb2cd64852d8db0796","name":"ks-out-V9_discrimination.out","bytes":1825},{"sha256":"4a996024bbbad8e1539a3edae7dbdba17f84d41cd7ee0a30c13090cfa81b22bf","name":"ks-validators-V9_discrimination.py","bytes":5900},{"sha256":"4da1c38772f23fa9d404e9ff1383bd32f24a2067c7a57c5cdf25f83ee6372746","name":"ksrun-check_record.md","bytes":2127},{"sha256":"4e13dde903e2e0633aca6934385e8468ad37afae51389f61b642d669321d162f","name":"ks-out-V1_tables.out","bytes":2521},{"sha256":"5205af45e29f0c3eb3962f8f169c5aa6162fc0699fe63ae051747b6a1468c54a","name":"ks-out-V5_model_spot.out","bytes":275},{"sha256":"53b4762a81962a0772b293b539e3a1262e9666012ce1f5c788987f9231ac3ab1","name":"ks-validators-V8_reduced_mc.py","bytes":2265},{"sha256":"573d6a7ca8b74cf3025d48f1aa7b4d1227f69bc7048c93df001c46d63a12bc33","name":"ks-validators-V4_fresh_primes.py","bytes":5395},{"sha256":"5a4a34ce3757d6602cd7f621c384a968b864d5bb18a1284fcaa009f8fb2ccdfa","name":"ksrun-ks_recount.cpp","bytes":2201},{"sha256":"5d8ef048b1020fc07ac4ade76dd3d07c42f29309ece4fce7f9c5ae87e590134a","name":"ks-out-registered-calib_score.out","bytes":1376},{"sha256":"5ee15f02801c73bcad76ed4c4dbaa7c171cc7723efd796552f0565e1e9d3ba9d","name":"ks-code-theory-secondorder_test.py","bytes":4285},{"sha256":"5f5a57f271cd4fea71c140b1aa675dbbf5f313c7e774673bda2e970c5af7db59","name":"ks-out-theory-compare.out","bytes":12420},{"sha256":"5f7c571de222b813ed528c4272955f7813b2a054d23e2a8ffc971c95c7ad3d4b","name":"ks-code-theory-lattice.py","bytes":9059},{"sha256":"60d4df6f2bed42fd227fd16cd7acc18eccd2774473cfac47f5cee14afe1c6b53","name":"ks-out-registered-calib.out","bytes":3619},{"sha256":"63cdf03225352d37ea9e79e03c786b17743e2b4ed5431351bfa8b79005f773c5","name":"ks-tables-step1-model_counts_L350.csv","bytes":1361012},{"sha256":"6e4e1f2cbb34b4c6c8c6a34c46b5fce97e77053cf8c41d793d557e002ba4189b","name":"ks-out-step1-compare_realL.out","bytes":1365},{"sha256":"7156b418b08e691e22c1b6a944d0fb60c1e6548d6fc88d578472ee49594f037b","name":"ks-tables-step1-model_counts_L120.csv","bytes":485512},{"sha256":"72ac6caafdc30a3c7ebac77cb945e5cbc50992166e3503d281b27de75b3031f1","name":"ks-out-V2_predictions.out","bytes":1380},{"sha256":"77c5429db3eff6dd3bc9d54a65dd8221c98288e77d97614b5615ac2f30afc078","name":"ks-out-theory-predict_W9699690.out","bytes":4028},{"sha256":"7cdbf357f50b76d3c0a1b2b4d297e177696b5f0d56d7ab1c5a64775b85acd31c","name":"ks-tables-step1-model_realL.csv","bytes":1186302},{"sha256":"7fb84e0fafa2435614d9a373b6a440eb888ce0f66316330613aa4b662658325c","name":"ks-frozen-rev0-real_counts.py","bytes":1613},{"sha256":"8195e74f9670e6fc2a50474be0f16769c611fc63e8c78629eb96705054ec7d8e","name":"ks-out-V7_used_heights.out","bytes":341},{"sha256":"828bc054b2bc4bf75e0c1c77f2d92e8d2f444426274ef9f74bc87b118d93c579","name":"ks-code-step1-akfit_real.py","bytes":1154},{"sha256":"829241a97f55344640c583cf5d2199165fdb664f36665bcee842aad5aa860661","name":"ks-tables-step1-model_counts_L90.csv","bytes":378512},{"sha256":"82a04c7d1992683bbb339682b73012a907518f435fec77beb108d20ad0ede3b6","name":"ks-validators-V0_frozen.py","bytes":6699},{"sha256":"842787acb4d732dc0b3ada0c325f6338b00b3165546b55419977194b7b31474f","name":"ks-kstep_bias_note.md","bytes":17358},{"sha256":"870496304c01928afb78266d56c9236a5f3d28c5b2303774e3ee01699dd2ea54","name":"ks-code-make_tables.py","bytes":8732},{"sha256":"8a02692f4d72226005fd651a1aa78e36b5ed5cc0365c58640f447cf96b6bdafc","name":"ks-SHA256SUMS.txt","bytes":8611},{"sha256":"8a6655e94e6d358fe5dbddb36f451f66e2714ee9e068dd77e7aa3ebfb24bc1b9","name":"ks-tables-registered-calib_counts.csv","bytes":2043301},{"sha256":"8e4d1ef6e52cb0bc1707001c64522ec755aea9e0008ad30e85237014dc7cd3f1","name":"ks-tables-step1-model_counts_L180.csv","bytes":728512},{"sha256":"93ac194eee060aaf9137a18d0cf43a5a4bad2f61b1847040e60afffccd49d9ab","name":"ks-frozen-prereg_rev1_utc.txt","bytes":21},{"sha256":"9696f87f8fdfbaecec5c0f044685458ea9410430957bb019ed522dc86bd4b14c","name":"ks-frozen-prereg_rev1.sha256.txt","bytes":157},{"sha256":"9a0d90a54e35277163d30f5bed31cfaba16573b21225dd6ee0a3156c19af3f9a","name":"ks-docs-prior_art_notes.md","bytes":6591},{"sha256":"9b02c0f5656db30ebfe70b2449e6465e05a3e1806933b44ac8c7d78aacd27892","name":"ks-out-registered-real_score.out","bytes":1287},{"sha256":"9d4e221207ce19e9d9538a129e1c5f479b5113c022cf0d183d94acf2fd11d65c","name":"ks-frozen-PREREG.md","bytes":4671},{"sha256":"a277536b95eba57f1914cf7dc3aa60c9142525ecbce9d69962ca11f44be41d36","name":"ks-tables-step1-model_counts_L30.csv","bytes":133712},{"sha256":"a30f111dc9154e97f32e2a544cb023531db480e1859de218d0e17ec82c31d3e8","name":"ks-code-registered-score.py","bytes":4145},{"sha256":"a52e19d87fd35f1d46a04084043bd8c7ef988e899e5dbfee415361774662b493","name":"ks-out-step1-lattice_try.out","bytes":1504},{"sha256":"a56e18e5c17f50be17300502ce6e59773a977fef7c1bec1f417420fe3b9efdad","name":"ks-code-step1-lattice_try.py","bytes":1311},{"sha256":"a5ca25e9a85dbdc088203fe72c5805adf9c0de6b478ea9b3ff1b272e5b891e4c","name":"ks-code-step1-akfit_model.py","bytes":2106},{"sha256":"a6aed4e24e67d69fb3a854671671223d25bb1a9c876909026e21c6bb70ba52ae","name":"ks-code-theory-THEORY_scrubbed.md","bytes":19434},{"sha256":"aadd3e4b4b03739fdaa2ca0c96f2643f00ed97acb57a4b2e256d4229d5c51b79","name":"ks-frozen-prereg.sha256.txt","bytes":1018},{"sha256":"afa5b785b7ed2b8136736aedb2f4d198ac22a36134f391c1ddecaea9398228e8","name":"ks-tables-manifest.json","bytes":47572},{"sha256":"b2781c2e34c0dfdb2f0076dbb61fa27f5036e1f57e004047157b7153e70a2e10","name":"ks-code-theory-predict.py","bytes":3736},{"sha256":"b2fd6b8085fe53262a1943fe1e532efd7c27d6a34faaa5a531e1b01da330f5b2","name":"ks-frozen-prereg_utc.txt","bytes":21},{"sha256":"b59445c06a83196e6ac859df6229ef0e231a0ef1423a4d7d25167eceba4009d3","name":"ks-code-step1-compare_realL.py","bytes":1429},{"sha256":"b9e56b1b2104a91f3614da3594281d2a023fad2906311d299f6e4e9d3f224c26","name":"ks-out-registered-real.out","bytes":973},{"sha256":"ba40218b483bfca92346a330486bbf277c62f39642991ca1d4e7deddba193d10","name":"ks-out-V4_fresh_primes.out","bytes":787},{"sha256":"bfffc7717728473c90468ae183e4cc52f85326b6792d6b36c274e25c1c88e12d","name":"ks-tables-theory-predictions_W9699690.csv","bytes":135043},{"sha256":"c6d41b53ff98fb0ca48fc519b4366551ea5e269303af80d4506dabb096f56cb1","name":"ks-tables-registered-fresh_predictions.csv","bytes":36544},{"sha256":"c8d9916b7af0edc601d5f3a0f33afc8156a5f4e5aa989b5519590b3dd0d4df9d","name":"ks-validators-V5_model_spot.py","bytes":2906},{"sha256":"cb0b5ed8edb276c0baefc6547b13d7cc71cc0db99b9dd665200b69aa4e7fea22","name":"ks-out-V8_reduced_mc.out","bytes":4010},{"sha256":"da49c7995c8615740d32a66c146110eb3ae0b371dbdb139e2044daef2c72b594","name":"ks-validators-V6_independent_score.py","bytes":6425},{"sha256":"db26d85cefed559646a4e458323dd713a8993e2f3c72f4c242891776e1015660","name":"ks-validators-V1_tables.py","bytes":8274},{"sha256":"db330356ac19597417a6d91dcb80a91082c03bcbd66d6250c20e74d50af27e7b","name":"ks-tables-step1-model_counts_L250.csv","bytes":974257},{"sha256":"dbbd2085068668939f119aa7379c2da56a79f6ce8281f8e8416c5963503d7f24","name":"ks-tables-step1-model_counts_L45.csv","bytes":198962},{"sha256":"df3ab45619b5945bef693e312d7094ba5506894ca4a3d4b9e10ada26f8b17738","name":"ks-tables-registered-real_counts_fresh.csv","bytes":129763},{"sha256":"e095e0edf2f4311dedea003ea645294c6deb4176ba4fee29c4189b10ba6f5e9e","name":"ks-validators-V7_used_heights.py","bytes":2592},{"sha256":"e3159fcbb6f5bf702a5689c3058070ad5f85461c6b63b27bb4ada8d94a8f8fb2","name":"ks-out-registered-out_calib.json","bytes":5107},{"sha256":"e6260158afffa8f927aa5a333764d458e9d30e2bdd48f91cf9c8ccb72f601fbd","name":"ks-tables-theory-predictions_W510510.csv","bytes":134879},{"sha256":"e72916f6744a85c7a41d3db46af67cb1fdbdc39b277418107615c51b767a7d97","name":"ksrun-ks_prereg.md","bytes":1809},{"sha256":"e79d574c41cc692df283f0b772abfa015ba6c9fa1b62ba81c2b89c31e6144934","name":"ks-docs-registered_test_plan.md","bytes":2803},{"sha256":"e897912e658725bd270047ed39741c4a5493dc03f1a0d0c6f5612c2d2d9fa7d8","name":"ks-out-V1_tables_originals.out","bytes":3277},{"sha256":"e90e7e4c43a2c0f1b1d87009d75bb4ff6dcea50e9f619d78090eb7cc94ef58a6","name":"ks-code-registered-fresh_predict.py","bytes":3122},{"sha256":"ebe6c643a621010be951b0c10932755422961950e1092ccddb4a73ce6fc1c0a5","name":"ks-tables-theory-predictions_2nd_W30.csv","bytes":60183},{"sha256":"ecaa62814ba41cdcda6c3da5c6527ec69b078072a8cd018caa78338ec968402a","name":"ks-code-registered-calib_model.py","bytes":2420},{"sha256":"efb3400d55096ec6d0fd46888a215e96d5239a8f9dde1c83dbf930ec7bb477ec","name":"ks-tables-step1-real7_counts.csv","bytes":113421},{"sha256":"f38890d76dee432c73f0a9df251e7ccbb7408f7438cbbddf5b49b0d1bb090f0a","name":"ks-tables-step1-real_counts.csv","bytes":162236},{"sha256":"f50fb3fbe2a601f2ac128c94666949913febaba41579fcf4432ffe84c82742ae","name":"ks-out-V3_score.out","bytes":487},{"sha256":"f6dd3b674bf0524b97944bd5e531f769f93a31e0ff9b0c91b5281824732b456e","name":"ks-out-theory-secondorder.out","bytes":7732},{"sha256":"fafd769e7d3c7d6756ebcf73c963d3c5d91b498c5614a118be7ee19a6fb44c6a","name":"ks-code-step1-analyze.py","bytes":4661},{"sha256":"fe4e7b7ae6c079773d536d9831a4cee92537b1844eb5faaea77d1a69cb2de8fc","name":"ks-tables-step1-model_counts_L60.csv","bytes":252512}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"report_sha256":"b7628dd14ee5f2b910f528a583924b1abdb0e0cd8159f730f3236c6eeaf52f5d","next_step_sha256":null,"research_authority":{"witness_status":null,"research_status":"pending","scopes":[]},"research_links":[],"duplicates":[],"cited_messages":[]}