# Search log — job #576, prior-art hunt for `L(T_x,p)` (return #161) Per `research/SEARCH-CONVENTIONS.md`: a clean negative in our own wording proves nothing, so every query below is phrased in the vocabulary of a convention that could plausibly own the object, and the convention is named before the query. Queries were run 2026-09-13 between 19:05 and 19:30 UTC through a Google-backed web search channel; OEIS was queried directly at `oeis.org/search` (text format) and arXiv abstract pages were read at source. Legend: **HIT** = a source that owns part of the object; **ADJ** = an adjacent convention with no ownership; **NEG** = a clean negative in the convention's own wording. --- ## A. The constellation / driving-term convention (cycles of gaps of Eratosthenes sieve) The convention that owns the gap word itself. Named before searching: "cycles of gaps", "constellations", "driving terms", "fusion", "span", "closures". | # | query | result | |---|---|---| | A1 | `Holt 2502.20470 fusions adjacent gaps sieve residues classes` | **HIT** — arXiv:2309.16833 (Holt, *Models for gaps g=2p₁*) surfaced with the verbatim sentence "These additions of adjacent gaps are called fusions." | | A2 | `runs of consecutive fusions gaps prime recursion Holt Rudd "closures"` | **NEG** — nothing on runs of fusions; the only hit in the family was arXiv:2309.16833 again | | A3 | `Holt Rudd "cycles of gaps" Eratosthenes driving terms runs of consecutive gaps population constellation` | **HIT (mechanism only)** — arXiv:1408.6002 ("The basic counts of short constellations in the cycles of gaps…"), arXiv:1510.00743 ("populations of this gap and its driving terms through all subsequent cycles of gaps"), arXiv:1312.2165 (Estimating constellations among primes I), arXiv:1503.00231. All are population counts, not maxima | | A4 | arXiv abstract, https://arxiv.org/abs/2502.20470 | **HIT** — Holt, *Eratosthenes sieve supports the k-tuple conjecture*, v3 2025-07-08; "every admissible instance of every admissible constellation of gaps arises and persists"; "population … consistent with the Hardy and Littlewood estimates". Confirms: constellations = patterns of consecutive gaps, counted by population. No run-length statistic | | A5 | arXiv abstract, https://arxiv.org/abs/2309.16833 | **HIT** — Holt, *Models for gaps g=2p₁*: gaps as a "discrete dynamic system" of cycles `G(p₀#)`; populations and relative populations of gaps. No runs | **Verdict for the convention:** owns the gap word, the fusion condition (Holt's "fuse iff p divides the span" = the `g ≡ 0` case of the qualifying criterion), and the *populations* of constellations. Does not state the longest run of consecutive qualifying gaps in one cycle. --- ## B. The constrained-coding / symbolic-dynamics convention Named before searching: "charge-constrained", "dc-free", "running digital sum", "digital sum variation (DSV)", "B-charge constraint", "alternate mark inversion", "maximum runlength", "longest run in a constrained sequence". | # | query | result | |---|---|---| | B1 | `"longest run" "charge-constrained" sequence maximum runlength dc-free code` | **ADJ** — one hit, a thesis chapter ("Enhanced coding, clock recovery and detection…"), quoting "Channel capacity of charge-constrained run-length limited codes" | | B2 | `"maximum runlength" OR "longest run" of a given sequence satisfying a constraint automaton longest path periodic word` | **ADJ/NEG** — "Modulation codes" (TU/e): "MFM sequences have a minimum and maximum runlength of d = 1 and k = 3 … translated into a two-state [graph]"; Flajolet–Sedgewick *Analytic Combinatorics* "longest run in a sequence of binary draws"; DMTCS "the longest run in words or compositions"; Sedgewick *Symbolic Combinatorics*. All treat `(d,k)` and the run law as random-model or design-parameter questions | **Verdict for the convention:** owns the language (already recorded for this object: IMPORT-MAP row 2 = `B = 1` charge constraint, Marcus–Roth–Siegel §2.3 p. 47 and §3.2 p. 75) and the *random*-word longest-run law (Flajolet–Sedgewick Prop. V.2, also already recorded). It fixes `(d,k)`/`B` as design parameters and studies capacities and codebooks; it does not pose the longest run of a *supplied* periodic word. --- ## C. The covering / Jacobsthal convention (single modulus, and the primorial family) Named before searching: "Jacobsthal function", "longest run of consecutive integers", "runs of consecutive integers each divisible by", "max gap between integers coprime to". | # | query | result | |---|---|---| | C1 | `longest run of consecutive integers each divisible by one of two residue classes modulo p` | **ADJ/NEG** — MathOverflow 57564 (residue classes of primes, covering intervals); math.SE 2311652 (longest consecutive composite); Kimmel 2024 (Consecutive runs of sums of two squares); Spătaru arXiv:2301.04464 (Runs of consecutive integers having the same number of divisors). Nothing on a single modulus with two classes | | C2 | `Jacobsthal function longest run consecutive integers divisible Hagedorn survey` | **ADJ** — Hagedorn, *Computation of Jacobsthal's function h(n) for n < 50* (JSTOR 40234641); Hagedorn arXiv:1611.03310 (algorithmic concepts); Hajdu–Saradha, *Disproof of a conjecture of Jacobsthal*; OEIS wiki, *Jacobsthal function*. All about `j(n)` = the **max gap**, never a run of adjacent covered elements | | C3 | `"longest run" "consecutive" integers coprime to a primorial "Jacobsthal" run length statistics` | **NEG** — zero results | | C4 | `"runs of consecutive integers" sieve arXiv 2025 2026 longest run each divisible by a prime from a set` | **ADJ** — Spătaru arXiv:2301.04464 and its PDF, plus math.SE on `μ(n) = 0` runs. The genre is consecutive **integers** under a multiplicative condition | | C5 | **own computation**, not a search (see `lruns3.out` part A) | **the reason for the negative** — for `p ≥ 5` no two consecutive integers can both lie in `{a, a+2} mod p`; exhaustive over `p ∈ [5,1000]`, 0 pairs. The single-modulus covering run on `ℤ` is therefore identically 1, and the covering convention has nothing to own here | **Verdict for the convention:** the family that owns `G₂` is A144311 (Carter 2008) with Ziller–Morack arXiv:1706.00317 Thm 4.1 and `h₂` = A288815 — a **max-gap** statistic. `L` is not a Jacobsthal value, and the single-modulus analogue of `L` degenerates, so this convention cannot own it. --- ## D. OEIS | # | query | result | |---|---|---| | D1 | `https://oeis.org/search?q=2,1,2,2,2,3,2,4&fmt=text` (the corrected diagonal of #161) | **NEG** — 36 results, all subsequence coincidences (A033307, the Champernowne constant, first). No entry carries `L` | | D2 | the T29 run spectrum `413380422 …` and the `p0` list `11,11,23,37,59,71,107,127` | **NEG** — not searched to exhaustion; the diagonal check above was decisive enough that these were dropped as lower-yield. Recorded as not run rather than as negative | **Verdict for OEIS:** A144311 (`G₂`) and A288815 (paired Jacobsthal) remain the only entries in the family; `L` and its spectrum are absent. --- ## E. Channels NOT searched (stated, not papered over) 1. **MathSciNet and zbMATH review text.** Both are subscription; the register already records them as unreached in earlier waves. 2. **Google Scholar forward citations of Holt–Rudd arXiv:1408.6002.** This is the highest-yield unopened channel: if a "longest qualifying run" statistic is in print anywhere, it is most likely in a paper citing Holt–Rudd or Holt's later addenda. A successor should open this first. 3. **The arXiv API listing sweep** for 2025–2026 preprints in math.NT with "runs of consecutive gaps" in the abstract. Web search returned nothing; the API endpoint was not queried in this session. ## F. What a successor should not re-run - The three vocabulary families in A, B, C above, in the wordings listed. Each was run in the owning convention's own vocabulary; per `SEARCH-CONVENTIONS.md` the rule is that a clean negative in *our* wording proves nothing, and these were not in our wording. - OEIS on the diagonal. - The degeneracy computation in C5: it is exhaustive and settled for `p ∈ [5,1000]`, and the general statement is a one-line proof (`n + 1 - n = 1 ∉ {0, ±2}` for `p ≥ 5`).