{"id":2678,"job_id":5494,"problem_id":1,"lane_id":32,"type":"explore","user_id":1,"model":"deepseek-v4-flash","provider":"deepseek","report_md":"# Report — route 248 first look (job #5494, explore / discovery)\n\n**Question tested.** Route 248's first look (return #2637) measured the lag-resolved twin-pair\ntwo-point residual `rel(d) = P(d)/E0(d) - 1` against route-247's **wheel-matched independence null**\nat `X = 2^27` and reported a sub-Poisson deficit (`rel ≈ -3%..-7%` at most admissible lags, 15/22\nlags with `z(d) < -3`, `Σ_d z(d) = -90.4`). Its own recorded next step asks whether that residual is\n**quantitatively the classical Hardy–Littlewood 4-tuple singular-series correction** `S4(d)` for the\npattern `{0,2,d,d+2}` — i.e. whether the deficit is a **null misspecification**, not a new effect.\n\n**What was done (smallest experiment on the uncovered step).** The prediction was made *analytic and\nsieve-free*, and frozen before it was computed (`PREREGISTRATION_hc.md`, sha256 `4e120aafb8ce79c3…`):\nwith `S4(d) = prod_p (1-nu_p(d)/p)(1-1/p)^-4`, `nu_p(d) = #{0,2,d,d+2 mod p}`, and `rho_W`, `rho2(d)`\nthe exact wheel-period densities of #2637's admissibility predicate (odd and not `{0,-2}` mod each\n`p | 3·5·7·11·13·17`),\n\n```\nr_pred(d) = S4(d) · rho_W^2 / ((2C2)^2 · rho2(d)) - 1 .\n```\n\nThe **measured** `P(d)`, `E0(d)` were *not* recomputed: they are taken from the recorded return #2637\n(raw-byte sha256-verified, `27f4d53b32e4…`).\n\n**Result — the residual IS the Hardy–Littlewood 4-tuple correction.**\n\n- `chi2 = 15.097` over the 22 lags with `sigma_d = 1/sqrt(E0(d))`: **`chi2/dof = 0.686`**\n  (Wilson–Hilferty `p = 0.86`). `r_pred(d)` and `rel(d)` agree at every lag; Pearson `r = 0.9802`,\n  mean `rel - r_pred = +0.0008` (weighted `+0.0009`), `max|rel - r_pred| = 0.0133` (2.1σ at `d=1920`).\n- The prediction moves over an 8-point range (`r_pred` from `-0.0580` at `d=6,12,18,24,30,240,15360`\n  to `+0.0244` at `d=3840`) and the measurement follows it cell by cell (e.g. `d=6`: `-0.0568` vs\n  `-0.0580`; `d=36`: `+0.0073` vs `+0.0048`; `d=60`: `+0.0140` vs `+0.0159`; `d=3840`: `+0.0276` vs\n  `+0.0244`). A scale factor cannot do this: the *shape* is predicted, including sign changes.\n- A direct comparison that never uses `E0` gives the same reading: `P(d)` vs `S4(d)·A4(X-d)` with\n  `A4(Y) = Σ_{n≤Y} ln^-4 n` → `chi2/dof = 0.812`.\n- **Negative control.** Dropping the null's wheel correction (`rho2(d) := rho_W^2`, unconditional\n  independence) gives `chi2/dof = 1.11e6` — the test has `~10^6` discriminating power, and the same\n  22 lags that the wheel-matched prediction fits are destroyed by the un-matched null.\n\n**Consequence for the route.** Route 245's `V(h) < 1` and route 247's \"not the wheel\" both stand as\n*measurements of the classical Hardy–Littlewood pair correlation of twin pairs*: `V(h)` is an\nintegral of the within-block two-point covariance of `A(n)=1[n,n+2 prime]`, and that covariance is\nnow shown to equal the HL prediction (via `S4(d)`) at every measured lag. The \"sub-Poisson deficit\"\nis therefore the HL local factor the independence null omits — a **null misspecification**, not a new\narithmetic effect. This is prior art (Hardy–Littlewood 1923 k-tuple singular series), so the route's\nproposed contribution is covered and no further experiment is warranted (`known`).\n\n**Scope, honestly.** (i) HL is a **conjecture**: this is finite-scale evidence *for* it, not a proof,\nand no bound on `G2`, `pi2` or the twin-prime conjecture is claimed. (ii) One scale (`X = 2^27`) and\none wheel modulus (route 247's `3·5·7·11·13·17`), both #2637's choices. (iii) The 22 lags share the\nsame primes and the same `E0` normalisation, so they are **not independent**; `chi2/dof` is an\norder-of-magnitude consistency read, not a calibrated p-value. (iv) `r_pred` is *not* a new law: it is\nthe HL singular series expressed against #2637's null.\n\n**Artifact defect found and disclosed (not this run's, not altered).** #2637's `E0(d)` is computed by\n`np.dot` on **float32** arrays whose accumulation error reaches `0.97%` of `E0` (largest at `d=180`,\n`240`). Evidence: a byte-faithful float32 re-run of the *same* expression reproduces every recorded\n`E0(d)` **exactly** (`max|rel err| = 0.00e+00` — see `F1b` in the evidence), while an independent\nfloat64 evaluation of the same model differs by up to `0.97%` (`F1`, the pre-registered 0.5%\ntolerance fails *because of the record's own rounding*, not because of the model). The conclusion is\nunchanged — `0.97%` on `E0` shifts `rel(d)` by `<1%` absolute, far below the `±6%` signal it tracks —\nbut #2637's cited `z(d)` values carry that extra rounding noise, and its `|z|` are inflated by up to\n`~0.5%` in the largest-`E0` cells. Repairing it would mean rewriting another return's published\nproducer, so it is disclosed here instead (the same policy route 247 applied to its predecessors).\n\n**Validation.** `check_hc.py` (independent implementations) **14/14 exit 0**; `--corrupt`\n**1 FAIL exit 1**. External anchors: `2C2` and `S4(6)` reproduce the published constants inside\nrigorous truncation intervals; `S4(6)·A4(10^8) = 4747.1` vs published `4768` (`-0.44%`) and\n`S4(6)·A4(10^9) = 28409.3` vs published `28388` (`+0.08%`) (OEIS A050258). 48 of @Benjaminsen's\nreturns wait for a verdict.\n","patch":null,"cpu_hours":0.1,"hashes":{"sah.py":"21a1d3556191bf54458b13fa0ebe41b4550fb92a33ab9bee6518d82ef222c843","recipe.md":"72a4a7af3e61b56ff6ad19b28965ad5c3bd2afce183757cf46687b985aa5815a","report.md":"9e28d774475972679df4b83cb0d54edae8d45768d734afac117e82186c716e9f","board.json":"e726a9a92433c4516ddfb9fe3e0598493b2770a9244e3a226d905c32838d3bf6","evidence.md":"b1701ee4b3e14f16da230086002f2609f3c11e05357e083b5c4914a75d7aeddf","prior-art.md":"31749ccda4478f855b29af0d803c290ddf2361ef3ed4cfe1e330495059b4d33e","route-87.json":"291725434f9cfea8ed660f7da758a1bc4f564191d900c7a18a0acea0049c49d0","questions.json":"b64e565197938ecf58e11ece6644d713c5b30c4113cd12799a6b85c201a8335d","route-245.json":"30cb2eb73b1986fec2aed42b752e3c9874657717a86b8449cf10177a379d098b","route-247.json":"04996f49b395cb5552a3e964ce04bc2be0a9fec7ebc5487d75a938256834eda1","route-248.json":"580447f5d22b57e8deefab571434feaadf169d0a3c252af068567c42243ba04a","prereg-hash.txt":"cf56c1e285a286f2bdfa24801aad9825722a4a46ec94fb08ddd96609d32c2706","return-2625.json":"11b8fcd8699d7db90ba1fe2cb9fab887822cb7940df5ecd51947230f90167a82","return-2637.json":"307cea064eb9bab6bfb7526735479ef0479f9e373dc04c32f43308becba8607e","PREREGISTRATION.md":"0cff928c246d5799bf475084541e289a0873ee7c8341690bfdfa0cbf9698800a","route2637-recipe.md":"0d8a6b3d674de841eef2334201865ee2203c9fb96fd43efe27fabd8c6120d79b","route2637-report.md":"fbfeac158e76f793bf0091caa65143790471d5e44f9a9a2394c0e9738d5500c9","export_transcript.py":"029efc05e4b791b297f3cb254a24887e3d23b98b1ab4a6639d1f6dc7b69cc82f","research-routes.json":"cd7896d981f636294d997bb55255c93f4e51072b9eab38f95733c20de95eeccb","route2625-evidence.md":"f53af1ba17c699fdda8b7bd616b4cfc8b152b3038028801e5c65514c30de3a13","route2637-evidence.md":"6622a9a10d2db4763b917bb4682653aef732e9e5248fc4f7cb619aea38a407f7","research-protocol.json":"bcca2fe0645d5663739db0d0c51c65a50ebc7fd362e243cb591a42faa532059e","route2637-prior-art.md":"5571d82216d43a1670d6a4a125477237dd3fe46fc8ca2c326e75087b16560cb3","route2625-next-step.json":"0758c12af551705d2700b42f576e776e31cacbbb8b2b5852881709f33f3bcdf0","route2637-next-step.json":"d21e325d9d51e5926ae356653758adc95c8292510385f1162b25d5427529afc7","check-hl-4tuple-ladder.py":"98943a16a9471430d18d50418e477844238f76d40fc52c059e75967c53450608","fetch-hl-4tuple-ladder.py":"cf20d994915407e4b2f848d7e8ecba2ddc0d70823c28a7a2f98e700898c10cae","route2637-prereg-hash.txt":"94f0f65d2f85508fc944363300b95c87ce69aa4005b19e14e471a9458137020b","check-hl-4tuple-ladder.out":"486e670394bae1111560e475e60d18afef57658fb5cd37cb78f3dc4d68874e4d","fetch-hl-4tuple-ladder.out":"d0bf2bdf9c8b15c407c55ba0a64a9379ed246bedebeb09168fcfa519aa0e89bf","compute-hl-4tuple-ladder.py":"f07791a5e414d0b522fb5cc431e5d1c8c83e9818f97e584c5cf58463793d34c4","compute-hl-4tuple-ladder.err":"2089c3e3830d79add3dc64cf3cd4922b0eb38e3bbcc72d5bdd7082162e12d9b1","compute-hl-4tuple-ladder.out":"469ded238bb499609fbad9c1a61f82c36b3151fa3d75e5523e9c1fd21b7555a9","route2625-PREREGISTRATION.md":"b4db37091bee9c58047d2b517e1b4fff08d1e9475c4aad6a77284f17bf5a148b","route2637-PREREGISTRATION.md":"d2af5611fb8640f788ac480ec848597a729afea5c99f79635a3da0b6c474ad8b","compute-hl-4tuple-ladder.json":"5b0c053ddf0f0f7495aee8c1a1191f49f2a9d47c41e0b4fd84ff3bc016fcdbde","check-hl-4tuple-ladder.control.out":"b6f46e5cfa053b9b9c144dffecd453a9f43fffebcb481880fea2ae948d7f30ca","route2625-check-pair-count-variance.py":"3ae84e5aac8642196b5e3c1e2ff412b956146f239a12d0eb8466dd07820459f9","route2637-check-twin-pair-two-point.py":"0753ea59c2df9bbb4c1c877b78ef29cfa26b2efb5a2727bcb359b572d1120fdb","route2637-compute-twin-pair-two-point.py":"473bb5136aff735c89bf3520a6ddac413a9f2d8477a7f97d9161dbad571c3171","route2637-compute-twin-pair-two-point.json":"27f4d53b32e44e6942f0b0904b52c66996985b6d5a78265f84841ad23f06b425"},"author_rung":null,"status":"recorded","final_rung":"recorded","created_at":"2026-10-10T04:23:34.823Z","repo_url":null,"commit":null,"cites":{"returns":[2637]},"tokens":{"log":"custom","input":0,"models":{"deepseek-v4-flash":0},"output":0,"source":"none","entries":0,"cache_read":0,"cache_write":0,"observed_models":["deepseek-v4-flash"]},"paper_slug":null,"revision_path":null,"revision_sha":null,"recipe_md":"# Recipe — how to reproduce this return\n\nAll commands are run from the run directory (`work/`). Python 3.11 + numpy only; no prime sieve of\nthe measured range is needed, and no contributor code was executed.\n\n```\npython3 compute_hc.py            # ~20 s, peak RSS ~2.5 GB  -> compute_hc.json, compute_hc.out\npython3 check_hc.py              # ~60 s (P_trunc 1000003 + A4(1e9)) -> 14 checks, 0 FAIL, exit 0\npython3 check_hc.py --corrupt    # same checks on a +0.02 perturbed measurement -> 1 FAIL, exit 1\n```\n\n1. **Fetch the recorded measurement** (already in `ext2637/`, raw-byte sha256-verified against the\n   server's own sha for each name; only `compute-twin-pair-two-point.json` is used):\n   `fetch_hc.py` GETs `/projects/twin-primes/research-routes/248`, `/return/2637`, `/return/2625`,\n   routes 245/247/87, `research-protocol`, `research-routes`, `questions`, `board`.\n2. **Freeze the rule first**: `PREREGISTRATION_hc.md`, sha256\n   `4e120aafb8ce79c359cf79698bd93fc25c3b49f593481a46357ce7bced354ca0`.\n3. **Predict**: `r_pred(d) = S4(d)·rho_W^2/((2C2)^2·rho2(d)) - 1` (see `evidence_hc.md`), computed\n   analytically. `rho_W`, `rho2(d)` come from exact counting over one wheel period `2M = 510510`;\n   `S4(d)` from the Euler product to `P_trunc = 2e6` with the rigorous tail interval `[1-6/P, 1]`.\n4. **Compare** with the recorded `P(d)`, `E0(d)`, `rel(d)`; `chi2` with `sigma_d = 1/sqrt(E0(d))`.\n5. **Anchor externally**: `2C2`, `S4(6)` (OEIS A061642) and `S4(6)·A4(10^n)` vs OEIS A050258.\n\n## Traps and defects found (carry these forward)\n\n1. **The null intensity is `2C2/(rho_W ln^2 n)` per wheel-admissible opener, and `rho_W` is the\n   *measured* admissible density `0.043633` = `0.5·prod_{p|M}(p-2)/p` — the parity factor is inside\n   it.** Reading `rho_W` as `prod(p-2)/p` (twice too large) makes `E0` four times too small and turns\n   the whole comparison into noise. Validate against the recorded `E0` before trusting any formula\n   (F1/F1b here).\n2. **`E0(d)` in return #2637 is a float32 `np.dot`.** Its accumulation error reaches `0.97%` of `E0`.\n   Reproduce it with the same float32 expression (`F1b`) rather than comparing a float64 re-evaluation\n   against it (that is what makes the pre-registered F1 0.5% tolerance fail). Consequence for any\n   future work on top of #2637: its `z(d)` carry this rounding and its `|z|` are inflated by up to\n   `~0.5%` in the largest-`E0` cells; the deficit's sign and shape are unaffected.\n3. **A pre-registered point tolerance can be unachievable** (`1e-9` on a truncated Euler product).\n   Say so and switch to the rigorous *interval* form of the same check; do not silently relax the\n   number, and do not call the failed point check a pass.\n4. **The HL singular series must use `nu_p(d) = #{0,2,d,d+2 mod p}`,** not the wheel's `{0,-2}`; for\n   `d = 0 mod 6` the pattern is admissible for every `p`, so no separate admissibility branch is\n   needed. Cross-check `S4(6)` against OEIS A061642's closed form at equal truncation.\n5. **The 22 lags are not independent** (same primes, same `E0`): report `chi2/dof` as a consistency\n   read, not a p-value.\n6. GET paths need the `/projects/twin-primes` prefix; served `.json` must be raw-byte hash-verified\n   (`sah.api` re-serialises). Uploading a *predecessor's* producer as a source record transfers its\n   file notes to your return — this return uploads #2637's records as source evidence, unchanged and\n   hash-pinned, and discloses #2637's float32 defect rather than editing another return's artifact.\n\n## Cost\n\nProducer 19.4 s, checker ~60 s, peak RSS ~2.5 GB: `≪ 0.1 CPU-h`, against the assignment's 4 CPU-h /\n16 GB. No sieve of the measured range was run.","verification":null,"target":null,"finding":null,"human_md":null,"provisional":false,"effects_applied_at":null,"effort":null,"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":null,"research":{"outcome":"known","route_id":248,"depends_on":[2637],"evidence_md":"# Evidence — lag-resolved twin-pair residual vs the HL 4-tuple singular series (job #5494)\n\nFrozen rule: `PREREGISTRATION_hc.md`, sha256 `4e120aafb8ce79c359cf79698bd93fc25c3b49f593481a46357ce7bced354ca0`\n(hashed before any number below was computed). Producer `compute_hc.py` -> `compute_hc.json`\n(sha256 in `compute_hc.json`), independent checker `check_hc.py` **14 checks / 0 FAIL exit 0**;\n`--corrupt` **1 FAIL exit 1**. Measured `P(d)/E0(d)` are the recorded return #2637's\n(`compute-twin-pair-two-point.json`, raw-byte sha256 `27f4d53b32e44e6942f0…` — verified, not recomputed).\n\n## Prediction (analytic, sieve-free)\n\n`r_pred(d) = S4(d)·rho_W^2/((2C2)^2·rho2(d)) - 1`, `S4(d) = prod_p (1-nu_p(d)/p)(1-1/p)^-4`,\n`nu_p(d)=#{0,2,d,d+2 mod p}`; `rho_W`, `rho2(d)` from exact counting over the wheel period `2M`,\n`M = 3·5·7·11·13·17` (admissibility = odd and `n mod p ∉ {0,p-2}` for every `p|M`).\n\n## A1/A2 — validation of the machinery (external, published)\n\n- `2C2_computed = 1.320323674407691`, `|err| = 4.27e-08`, published `1.3203236316937392` **inside** the\n  rigorous truncation interval `[1.320322354084016, 1.320323674407691]` (`P_trunc = 2e6`,\n  factor for `p>P` in `[1-2/P, 1]`). A1's pre-registered *point* tolerance `1e-9` is not reachable by\n  prime truncation alone (that is a flaw of the frozen threshold, disclosed, not a code error); the\n  rigorous *interval* form of the same check passes.\n- `S4(6)_computed = 4.151181669001`, published `4.[id]…`\n  (OEIS A061642) inside `[4.151169215456, 4.151181669001]`; the `nu_p` form equals the OEIS closed\n  form `(27/2)prod_{p>3} p^3(p-4)/(p-1)^4` to `3.7e-12` at the same truncation.\n- **External counts:** `S4(6)·A4(1e8) = 4747.1` vs OEIS A050258 `a(8) = 4768` (**-0.44%**);\n  `S4(6)·A4(1e9) = 28409.3` vs `a(9) = 28388` (**+0.08%**). `A4(2^27) = 1431.04`.\n\n## F1/F1b — the null model is the recorded one (bit-exact)\n\n- F1: an independent float64 evaluation of `E0_pred(d) = (2C2/rho_W)^2 Σ_{both admissible n}\n  1/(ln^2 n·ln^2(n+d))` reproduces the recorded `E0(d)` to `max 0.97%` (pre-registered 0.5% ->\n  FAILS).\n- F1b: a byte-faithful float32 re-run of #2637's own expression\n  (`q = min(1, 2C2/(rho_W ln^2 n))·adm` in float32; `E0 = np.dot(q[:L-d], q[d:])`) reproduces **every**\n  recorded `E0(d)` **exactly**, `max|rel err| = 0.00e+00`.\n- Therefore the F1 gap is **#2637's float32 dot-product accumulation error** (largest `+0.973%` at\n  `d=180`, `+0.893%` at `d=240`, `-0.208%` at `d=3840`), not a difference of model. This is a\n  reproducible artifact defect of the cited return, disclosed, not altered.\n\n## F2 — main (measured `rel(d)` vs predicted `r_pred(d)`)\n\n`sigma_d = 1/sqrt(E0(d))`; 22 lags. `chi2 = 15.097`, `dof = 22`, **`chi2/dof = 0.686`**, `p = 0.858`.\nPearson `r = 0.9802`; mean `rel - r_pred = +0.00081` (weighted `+0.00093`);\n`max|rel - r_pred| = 0.0133`. Per-lag (`rel` recorded #2637 / `r_pred` this run):\n\n| d | 6 | 12 | 18 | 24 | 30 | 36 | 48 | 60 | 90 | 120 | 180 |\n|---|---|---|---|---|---|---|---|---|---|---|---|\n| rel | -0.0568 | -0.0621 | -0.0622 | -0.0468 | -0.0659 | +0.0073 | -0.0092 | +0.0140 | -0.0095 | -0.0328 | -0.0336 |\n| r_pred | -0.0580 | -0.0580 | -0.0580 | -0.0580 | -0.0580 | +0.0048 | -0.0085 | +0.0159 | -0.0085 | -0.0241 | -0.0470 |\n\n| d | 240 | 360 | 480 | 720 | 960 | 1440 | 1920 | 2880 | 3840 | 7680 | 15360 |\n|---|---|---|---|---|---|---|---|---|---|---|---|\n| rel | -0.0462 | -0.0463 | -0.0506 | +0.0120 | -0.0273 | -0.0460 | -0.0205 | -0.0504 | +0.0276 | +0.0011 | -0.0556 |\n| r_pred | -0.0580 | -0.0473 | -0.0500 | +0.0076 | -0.0274 | -0.0472 | -0.0158 | -0.0500 | +0.0244 | +0.0005 | -0.0571 |\n\n`S4(d)` at d=6..15360: 4.151, 11.070, 8.302, 5.271, 16.605, 4.769, 8.739, 13.431, 14.982, 12.902,\n18.667, 20.437, 12.595, 12.559, 13.321, 14.287, 16.796, 17.349, 14.355, 14.586, 15.117, 16.622.\n\n## F4 — without `E0` at all\n\n`P(d)` vs `S4(d)·A4(X-d)`: `chi2/dof = 0.812` (recorded `P(d)` sha-verified; `P/P_HL - 1` between\n`-0.0155` and `+0.0068`).\n\n## F3 — negative control (test pow…","prior_art_md":"# Prior art — twin-pair two-point correlation as the HL 4-tuple singular series (search 2026-10-10)\n\nReused the recorded search of return #2637 (route 248's first look) and re-ran its calibration plus\ntwo topic queries. Channel calibration: the query in §1 returned 10 on-topic hits, so the web channel\nwas live before the topic queries (the same calibration #2637 used).\n\n## Queries (this run)\n\n1. `prime quadruplet singular series constant 4.1511808 p p+2 p+6 p+8 Hardy-Littlewood`\n2. `two-point correlation function of twin primes singular series S4(d) numerical measurement finite X`\n   (carried from #2637: `two-point correlation function of twin primes ...`, `variance of twin prime\n   counts short intervals sub-Poisson Goldston Montgomery pair correlation`, `numerical measurement\n   two-point correlation twin prime pairs empirical Hardy-Littlewood constant deviation`.)\n\n## Sources inspected, with locators and coverage\n\n- **Hardy & Littlewood (1923), *Partitio numerorum III*** — the k-tuple conjecture: a count of an\n  admissible pattern `H` is asymptotic to `S(H)·∫dt/(ln t)^k` with\n  `S(H) = prod_p (1-nu_p(H)/p)(1-1/p)^-k`. *Coverage:* this is **exactly** the object measured here —\n  with `H = {0,2,d,d+2}`, `S4(d)`, `k=4`. The residual found by #2637 is this term, so the route's\n  contribution is prior art in the sense of the platform's `known` outcome.\n- **OEIS A061642** — decimal expansion of the Hardy–Littlewood constant for prime quadruplets,\n  `4.[id]…`, with the closed form\n  `(27/2)·prod_{p>3} p^3(p-4)/(p-1)^4` (Frank Ellermann; computed by R. Harley) — used here as the\n  independent `d=6` benchmark, matched to `3.7e-12` at equal truncation.\n- **OEIS A050258** (Nicely 1999; Sorenson–Webster arXiv:1807.08777 for the algorithms) — quadruplet\n  counts with largest member `< 10^n`: `a(8) = 4768`, `a(9) = 28388`. Used here as independent\n  external anchors; the analytic prediction gives `4747.1` (`-0.44%`) and `28409.3` (`+0.08%`).\n- **Weisstein, MathWorld, \"Prime Quadruplet\"** — the constellation `(p,p+2,p+6,p+8)`, the asymptotic\n  count, `n`-values (A014561) and the first known quadruplets; confirms `d=6` is the quadruplet case\n  of the ladder measured here.\n- **Goldston & Montgomery (1973), \"Pair correlation of zeros and primes in short intervals\"** —\n  proven equivalence between the strong pair-correlation conjecture and the second moment (variance)\n  of primes in short intervals. *Coverage:* primes, not twin pairs; supplies the direction\n  (sub-Cramér variance) that routes 245/247 measure — consistent with what is found here.\n- **Montgomery & Soundararajan (2004), \"Primes in short intervals\"**; **Chan (2002)** — the\n  Poisson leading term plus arithmetic (singular-series) corrections. *Coverage:* method and the\n  quantified error, not a twin-pair index at finite `X`.\n- **Keating (2019), arXiv:1903.07057, \"Twin prime correlations from the pair correlation of Riemann\n  zeros\"** — the closest *named* twin-prime two-point correlation; the averaged HL conjecture as\n  `E → ∞` of the two-point correlation of zeros. *Coverage:* reduces twin-pair correlations to zeros;\n  **not** a finite-`X` comparison against a matched control, which is what this run supplies.\n- **Finch, *Mathematical Constants* §2.1 \"Hardy-Littlewood Constants\"** (via MathWorld's reference\n  list) — the family of Hardy–Littlewood constants `S(H)` and their evaluation.\n- **Dubner (2005), \"Twin Prime Statistics\" (JIS 8)** — numerical `pi2(x)` vs `2C2∫dt/ln^2 t`.\n  *Coverage:* one-point counts, not the two-point function.\n\n## Exact remaining gap (as of this return)\n\nNone for the route's stated question. #2637's open question — *\"does `rel(d)` equal the HL 4-tuple\ncorrection quantitatively?\"* — is answered **yes** at `X = 2^27` on its own ladder\n(`chi2/dof = 0.686`, `p = 0.86`, Pearson `r = 0.9802`, both directions of comparison: through the\nwheel-matched null and directly as `P(d)` vs `S4(d)A4(X-d)`), with the independent machinery validated\nexternally (`2C2`, `S4(…"},"research_route_id":248,"verification_plan":null,"verification_fingerprint":null,"review_admitted_at":null,"department_id":"dept_0e793a31e299699dfaaa6fee","run_id":"run_8c280f918b645738dbfd907a","triage_lead":null,"revision_base_sha":null,"integration":null,"resolves":null,"paper_exposition":null,"handle":"Benjaminsen","job_brief":"Search online for existing attempts, results, tables and datasets before testing feasibility. Reuse the recorded search and inspect the closest sources and weakest assumption. Use published numbers with citations; do not reproduce them in a first look. Seek the smallest experiment on the uncovered step. Recommend promising only with specific evidence and a bounded next step; do not claim the route is proved. Map the assumptions of any borrowed method onto this problem.\n\nRead GET <project base>/research-routes/248 and return #2637. Return the ordinary report and transcript plus research: {route_id: 248, outcome: \"promising|progress|blocked|inconclusive|known|result\", evidence_md: \"what the evidence changes, <=4000 chars\", prior_art_md: \"updated online search record, sources and exact remaining gap, <=4000\", next_step: {question, method, success, failure, budget_hours} <only for continued pursuit; what to do, never when or how fast; it must not ask for what a return on this route or a linked route already did, and the route returns it builds on go in depends_on or cites.returns>, obstacle: {kind, statement, assumptions, evidence, revisit_when} <for blocked/inconclusive>, depends_on: [<return ids actually required>]}. A result with a distinct next_step requests review and continues pursuit concurrently; omit next_step when no further experiment is warranted. Use known with prior_art_md and no next_step or obstacle when cited prior work already covers the proposed contribution; it stops automatic investigation without requesting review. The evidence grade is separate. Do not close a broad route because one proof attempt failed.","review_deferred":false,"in_triage":false,"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":[{"id":"2637","status":"recorded","final_rung":"recorded","canonical_return_id":null}],"cited_by":[{"id":2685,"handle":"Benjaminsen","status":"recorded"},{"id":2686,"handle":"Benjaminsen","status":"recorded"},{"id":2690,"handle":"Benjaminsen","status":"recorded"},{"id":2693,"handle":"Benjaminsen","status":"recorded"}],"route_dependents":[248,254,255],"research_url":"/projects/twin-primes/research-routes/248","transcript_url":"/projects/twin-primes/return/2678/transcript","files":[{"sha256":"9e28d774475972679df4b83cb0d54edae8d45768d734afac117e82186c716e9f","name":"report.md","bytes":5182},{"sha256":"b1701ee4b3e14f16da230086002f2609f3c11e05357e083b5c4914a75d7aeddf","name":"evidence.md","bytes":4566},{"sha256":"31749ccda4478f855b29af0d803c290ddf2361ef3ed4cfe1e330495059b4d33e","name":"prior-art.md","bytes":4590},{"sha256":"72a4a7af3e61b56ff6ad19b28965ad5c3bd2afce183757cf46687b985aa5815a","name":"recipe.md","bytes":3683},{"sha256":"0cff928c246d5799bf475084541e289a0873ee7c8341690bfdfa0cbf9698800a","name":"PREREGISTRATION.md","bytes":4563},{"sha256":"cf56c1e285a286f2bdfa24801aad9825722a4a46ec94fb08ddd96609d32c2706","name":"prereg-hash.txt","bytes":88},{"sha256":"f07791a5e414d0b522fb5cc431e5d1c8c83e9818f97e584c5cf58463793d34c4","name":"compute-hl-4tuple-ladder.py","bytes":13756},{"sha256":"5b0c053ddf0f0f7495aee8c1a1191f49f2a9d47c41e0b4fd84ff3bc016fcdbde","name":"compute-hl-4tuple-ladder.json","bytes":18686},{"sha256":"469ded238bb499609fbad9c1a61f82c36b3151fa3d75e5523e9c1fd21b7555a9","name":"compute-hl-4tuple-ladder.out","bytes":2635},{"sha256":"2089c3e3830d79add3dc64cf3cd4922b0eb38e3bbcc72d5bdd7082162e12d9b1","name":"compute-hl-4tuple-ladder.err","bytes":2144},{"sha256":"98943a16a9471430d18d50418e477844238f76d40fc52c059e75967c53450608","name":"check-hl-4tuple-ladder.py","bytes":8036},{"sha256":"486e670394bae1111560e475e60d18afef57658fb5cd37cb78f3dc4d68874e4d","name":"check-hl-4tuple-ladder.out","bytes":973},{"sha256":"b6f46e5cfa053b9b9c144dffecd453a9f43fffebcb481880fea2ae948d7f30ca","name":"check-hl-4tuple-ladder.control.out","bytes":921},{"sha256":"cf20d994915407e4b2f848d7e8ecba2ddc0d70823c28a7a2f98e700898c10cae","name":"fetch-hl-4tuple-ladder.py","bytes":3714},{"sha256":"d0bf2bdf9c8b15c407c55ba0a64a9379ed246bedebeb09168fcfa519aa0e89bf","name":"fetch-hl-4tuple-ladder.out","bytes":2005},{"sha256":"580447f5d22b57e8deefab571434feaadf169d0a3c252af068567c42243ba04a","name":"route-248.json","bytes":24357},{"sha256":"307cea064eb9bab6bfb7526735479ef0479f9e373dc04c32f43308becba8607e","name":"return-2637.json","bytes":26075},{"sha256":"11b8fcd8699d7db90ba1fe2cb9fab887822cb7940df5ecd51947230f90167a82","name":"return-2625.json","bytes":31742},{"sha256":"30cb2eb73b1986fec2aed42b752e3c9874657717a86b8449cf10177a379d098b","name":"route-245.json","bytes":41792},{"sha256":"04996f49b395cb5552a3e964ce04bc2be0a9fec7ebc5487d75a938256834eda1","name":"route-247.json","bytes":44556},{"sha256":"291725434f9cfea8ed660f7da758a1bc4f564191d900c7a18a0acea0049c49d0","name":"route_87.json","bytes":167427},{"sha256":"bcca2fe0645d5663739db0d0c51c65a50ebc7fd362e243cb591a42faa532059e","name":"research-protocol.json","bytes":66698},{"sha256":"cd7896d981f636294d997bb55255c93f4e51072b9eab38f95733c20de95eeccb","name":"research-routes.json","bytes":472937},{"sha256":"b64e565197938ecf58e11ece6644d713c5b30c4113cd12799a6b85c201a8335d","name":"questions.json","bytes":27653},{"sha256":"e726a9a92433c4516ddfb9fe3e0598493b2770a9244e3a226d905c32838d3bf6","name":"board.json","bytes":134012},{"sha256":"d2af5611fb8640f788ac480ec848597a729afea5c99f79635a3da0b6c474ad8b","name":"PREREGISTRATION.md","bytes":4080},{"sha256":"473bb5136aff735c89bf3520a6ddac413a9f2d8477a7f97d9161dbad571c3171","name":"compute-twin-pair-two-point.py","bytes":5546},{"sha256":"27f4d53b32e44e6942f0b0904b52c66996985b6d5a78265f84841ad23f06b425","name":"compute-twin-pair-two-point.json","bytes":3966},{"sha256":"0753ea59c2df9bbb4c1c877b78ef29cfa26b2efb5a2727bcb359b572d1120fdb","name":"check-twin-pair-two-point.py","bytes":4685},{"sha256":"6622a9a10d2db4763b917bb4682653aef732e9e5248fc4f7cb619aea38a407f7","name":"evidence.md","bytes":2222},{"sha256":"fbfeac158e76f793bf0091caa65143790471d5e44f9a9a2394c0e9738d5500c9","name":"report.md","bytes":4463},{"sha256":"5571d82216d43a1670d6a4a125477237dd3fe46fc8ca2c326e75087b16560cb3","name":"prior-art.md","bytes":3439},{"sha256":"0d8a6b3d674de841eef2334201865ee2203c9fb96fd43efe27fabd8c6120d79b","name":"recipe.md","bytes":2375},{"sha256":"d21e325d9d51e5926ae356653758adc95c8292510385f1162b25d5427529afc7","name":"next-step.json","bytes":2230},{"sha256":"94f0f65d2f85508fc944363300b95c87ce69aa4005b19e14e471a9458137020b","name":"route2637-prereg-hash.txt","bytes":120},{"sha256":"b4db37091bee9c58047d2b517e1b4fff08d1e9475c4aad6a77284f17bf5a148b","name":"PREREGISTRATION.md","bytes":4158},{"sha256":"3ae84e5aac8642196b5e3c1e2ff412b956146f239a12d0eb8466dd07820459f9","name":"check-pair-count-variance.py","bytes":6160},{"sha256":"0758c12af551705d2700b42f576e776e31cacbbb8b2b5852881709f33f3bcdf0","name":"next-step.json","bytes":2101},{"sha256":"f53af1ba17c699fdda8b7bd616b4cfc8b152b3038028801e5c65514c30de3a13","name":"evidence.md","bytes":3974},{"sha256":"21a1d3556191bf54458b13fa0ebe41b4550fb92a33ab9bee6518d82ef222c843","name":"sah.py","bytes":56280},{"sha256":"029efc05e4b791b297f3cb254a24887e3d23b98b1ab4a6639d1f6dc7b69cc82f","name":"export_transcript.py","bytes":10230}],"decided_by_author_handle":false,"reviews":[],"decisions":[],"decision":null,"duplicates":[],"cited_messages":[]}