# The lit/imports corrections reach the registries: three import rows written in, three search-convention rows added, four other files corrected

<!-- ledger
id: Q-applied-0828-registries
status: ANSWERED
todo: Z5, 4 (retired)
question: Which lit/imports corrections reached the registries?
verdict: 13 edits across six files, plus one appended changelog section: IMPORT-MAP gains rows 15 (no THEOREM, the column struck), 16 (STALE) and 17 (CLOSURE, two-sided) with its three counts recounted to seventeen rows; SEARCH-CONVENTIONS gains the Kourbatov-b, generalized-Dickman and VC-of-multiples rows with each channel's calibration state; import-map-construction's determinantal cell, natal-cap-36's separate-census sentence, one REFUTED row and one PRIOR-ART block are corrected; GLOSSARY, TODO, QUESTIONS, the five loose hyperuniformity documents and two staging notes are left for their holders and listed here.
-->

*(2026-08-28. Integration pass over `applied-0828-litimports.md` §5, and the
notes it cites. Fence: this pass edited `research/IMPORT-MAP.md`,
`research/SEARCH-CONVENTIONS.md`, `research/history/staging/import-map-construction.md`
§1, `research/natal-cap-36-skeleton-door.md`, `research/REFUTED.md`,
`research/PRIOR-ART.md` and `research/history/CHANGELOG.md` (appended), and
wrote only this file. It edited no `.js`, no `paper/` file, no TODO.md, no
README.md and not `research/GLOSSARY.md`, which another holder owns this
session. It ran no git command and did not run `research/qc.js`. Every number
below is carried from the source notes, none is recomputed here, and the source
notes remain HELD.)*

---

## 0. What is still open after this pass

**Nothing here was re-verified.** Every reading applied below is inherited from
`redteam-0828-litimports.md` and the six notes `applied-0828-litimports.md`
corrected. The load-bearing one, that Saffari-Vaughan II Theorem 10 reads
`x^{6/11+ε} < y ⩽ x`, still rests on one 150 dpi rendering of one scan of
printed p. 7. If that reading is wrong, row 15's struck THEOREM column comes
back and the `natal-cap-36-skeleton-door.md` sentence edited below is wrong in
the conservative direction.

**The gate has not run and one guard is knowingly out of compliance.** This
file's ledger block names TODO items Z5 and 4, and the `Ledger:` lines of those
items list `Q-redteam-0828-litimports` but not `Q-applied-0828-registries`.
TODO.md is outside the fence, so the ledger guard will flag it until a holder of
TODO.md adds the id. `research/QUESTIONS.md` is generated and is stale for this
id and for `Q-record-deficit` until `node research/qc.js --index` runs.

**Two sources of the `search-convention` check changed.** §1 of
`SEARCH-CONVENTIONS.md` is parsed at run time and its rows set what clears the
check, so three added rows change what the gate accepts. That is the intended
effect of a convention row and it is also the hazard the file names in its §6;
the three rows added here each name a convention the literature actually uses,
and each carries the calibration state of its own channels.

**Four of the nine rows `applied-0828-litimports.md` §5a collected are not
applied here, and its `SEARCH-CONVENTIONS.md` row is applied in part**: of the
five convention rows proposed across the source notes, the Kourbatov-`b`,
generalized-Dickman and VC rows are in, and the fractional-parts and
sieved-set-instance rows are not. All of it is listed in §3 with reasons.

---

## 1. The edits, by file

### 1a. `research/IMPORT-MAP.md`, 6 edits

| section | old reading | new reading |
|---|---|---|
| §0a, the regrade rule | the map is SPENT as a queue after thirteen original rows, with the standing rule for anything new | the same, plus the four rows added under that rule and their outcomes: row 14 landed 2026-08-21, rows 15 and 17 landed 2026-08-28, row 16 is STALE because it was priced as an experiment two in-house notes had already executed |
| §2, preamble | "All thirteen rows have now run; the last, row 13, landed 2026-08-20" | all seventeen rows have run or been overtaken, with the four dates and the reason row 16 is stale |
| §2, table, row 15 | absent (the table ended at row 14) | the fractional parts of `x/n` and `x/p`, EXACT-IDENTITY at the marginal face, SPLIT circularity, payoff PUBLISHED-ANCHOR + WALL-ADDRESS with **no THEOREM**, status LANDED 2026-08-28, no route, prediction HIT. The status cell carries the covered-branch measurements (0.20 / 0.70 / 5.68 / 1.10% at @13 to @23), the strike (`W^{1/12} < M_T ≤ √W`, so no fixed-`M` asymptotic and no cap-36 upgrade), and the sharper wall (`M_T > lB` is exactly `q > W/M_T`, so the phase never wraps) |
| §2, table, row 16 | absent | the theory of records, STRONG-ANALOGY, CLEAN, payoff WALL-ADDRESS only, status STALE, naming both notes that executed it and Kourbatov's published `b` as the prior-art half, with the custody rider that neither note's figures may leave it |
| §2, table, row 17 | absent | repulsive point processes, EXACT-IDENTITY at the statistic and NEGATIVE at the model, CLEAN, payoff CLOSURE + PUBLISHED-ANCHOR, status LANDED 2026-08-28, carrying the two-sided closure (`g` is 0 or at least 2.3812, `g(6) = 2.661728` at `x = 11`, `d = 6` the exact minimiser), the withdrawn Gibbs and Coulomb clause, and the demotion of the hyperuniformity reading to sub-Poisson along the `L = y^u` family |
| §2, Counts | "Fourteen rows, all fourteen resolved ... five EXACT-IDENTITY, nine STRONG-ANALOGY ... ten CLEAN ... three THEOREM, four DERIVED-CONSTANT, seven WALL-ADDRESS, three PUBLISHED-ANCHOR, three CLOSURE" | seventeen rows, none UNTRIED; seven EXACT-IDENTITY, ten STRONG-ANALOGY, zero VOCABULARY-ONLY (twelve rejected at that grade, one of them, determinantal point processes, promoted to row 17); twelve CLEAN, three TPC-STRENGTH, rows 2, 4 and 15 split; priced four THEOREM, five DERIVED-CONSTANT, nine WALL-ADDRESS, five PUBLISHED-ANCHOR, four CLOSURE, with row 15's lost THEOREM and row 16's banked wall address added to the priced-against-banked list |

The rows are `applied-0828-litimports.md` §5b's texts. The only additions are the
status cells for rows 15 and 17, condensed from `import-fracparts.md` §8 and
`import-repulsive.md` §7 so each row states its own outcome, as every other row
in that table does. No §3 paragraph was written for rows 15 to 17; §3 stops at
row 13 and row 14 already has none.

### 1b. `research/SEARCH-CONVENTIONS.md` §1, 3 rows added

| row | what it now says |
|---|---|
| the record-location deficit | owning convention is Kourbatov's `b` in `E_1 = a log(p/a) − ba`, or the Gumbel mode `µ*` of the standardized record gaps; `b ≈ 2/k` by default, `b ≈ 1.2597` and `µ* = −1.659` for twins below `10^15`; search `b`, `E_1`, `median-unbiased`, `standardized maximal gaps`, never "location parameter", "trend load" or "record deficit". Sources read at page image; the house side's figures are outside output custody and stay in their file |
| the two-class sifted short-interval variance | at `θ = 1` the convention is "the variance of integers without small prime factors in short intervals" with limit law `GD(θ)`; at `θ ≥ 2` it is "the distribution of `k`-tuples of reduced residues", statistic `M_k^D(q,h)`. Carries the scope warning that Gorodetsky is `κ = 1` only, confirmed at source, so his `λ` is `λ₁` and is not the comparand for a two-class census, and that at `θ = 2` only Aryan's upper bound is in print. Channel state recorded: absent on six channels calibrated in session, owed on the arXiv API leg (HTTP 429 all session), Google Scholar, MathSciNet review text and *Opera de Cribro* §6.10 Prop. 6.26 |
| the VC dimension of the two-class range space | owning conventions are Thomas's "the VC-dimension of a class of multiples of the primes" and Helmbold-Sloan-Warmuth's "learning integer lattices"; our own wording returns nothing and no number-theory-side convention has been located. Channel state recorded: Thomas read in full in both versions and DISJOINT at theorem level, forward citations zero on three channels each calibrated in session, and HSW Theorem 3.1 **OWED**, closed access, abstract only, its `k = 1` multiples-of-`d` reading single-source through Thomas |

The Kourbatov row is `lit-kourbatov-shortfall.md` §7's draft. The variance row is
`lit-dickman-variance.md` §6.1's first draft row with the `κ = 1` scope warning
and the channel state folded in. The VC row is written here from
`lit-vc-multiples.md` §§0, 1 and 5, which drafted no row.

### 1c. Four smaller edits

| file | place | old reading | new reading |
|---|---|---|---|
| `research/history/staging/import-map-construction.md` | §1, the determinantal row's "why it is not a fit" cell | pair density at distance 2 equal to `∏(1−2/p)/∏(1−1/p)² > 1`, positively correlated | the two-point inequality `K_ii K_jj − \|K_ij\|² ≤ P(i)P(j)`; the twin tile fails at `d = 6`, its first admissible separation, with `g(6) = 6∏_{5≤p≤x}(1−4/(p−2)²) = 2.661728` at `x = 11` and never below 2.3812, while `g(2) = 0` identically since mod 3 forces `6 \| d`; the one-class analogue fails at `d = 2` with `g₁(2) = 2∏_{2<p≤x}(1−1/(p−1)²) = 1.3535` at `x = 11`; the conclusion is unchanged and the field re-entered the map as row 17. The old product evaluates to 0 at every level, its `p = 2` factor being `1 − 2/2` |
| `research/natal-cap-36-skeleton-door.md` | Measurement D, the separate-census sentence | the census stands with no rung and no scope | the census is MEASURED at its nine finite levels and nothing upgrades it to PROVEN, for three named reasons: Theorem 10 holds only on `W^{1/12} < M ≤ √W`, its `exp(−C(log W/log log W)^{1/3})` saving certifies no finite level, and it controls one marginal while the door is joint |
| `research/REFUTED.md` | table, one row added | absent | modelling the twin tile as repulsive (Hermitian determinantal, negatively-associated, strongly Rayleigh) or as attractive (permanental, positively-associated) is CLOSED in both signs, because `g` is 0 or at least 2.3812 and nothing between |
| `research/PRIOR-ART.md` | the Kourbatov-Wolf bullet | records the 2019 successor's framework only | adds that the 2013 predecessor already owns the location statistic: `b ≈ 1.2597` median-unbiased below `10^15`, `−b = µ* + γ` with `µ* = −1.659`, the repo's record-location deficit is that `b`, and the deficit is documented in print and explained in neither corpus |

`research/history/CHANGELOG.md` carries one appended section, one paragraph per
file above, naming the records each correction came from.

---

## 2. What is not claimed

No route opened, no exponent moved, no measurement changed. Rows 15 and 17 are
the fifteenth and sixteenth landings on this map and the map's route count stays
at zero. Row 15 banks less than it was priced at, row 16 banks a wall address
that two other notes had already earned, and row 17 closes a family of models
nobody in this corpus had built. The `natal-cap-36` edit lowers a claim rather
than raising one.

---

## 3. Left for their holders

- **`research/GLOSSARY.md`, the Hyperuniformity entry** (line 393): "the natal
  field's suppressed large-scale density fluctuation: window counts are
  sub-Poisson, `Var/E` drifting 0.152 to 0.396 across @7..@37 rather than
  sitting at 1". That is sub-Poisson, not hyperuniformity in Torquato's sense,
  where the classification is by the large-`R` asymptotics of one configuration:
  every fixed level here is class I, vacuously, by periodicity. The entry should
  say sub-Poisson or split in two. Not edited, another holder owns that file
  this session. Source: `import-repulsive.md` §3.
- **The same loose use of the word in five places**:
  `research/level-ledger-tight.md`, `research/FOLD-PROFILE.md` §3,
  `research/discrepancy-two-class.md`, `research/sift-limit-attack.md`,
  `research/natal-cap-29-sigma-plateau.js`. Outside the fence.
- **`SEARCH-CONVENTIONS.md`, two further proposed rows**: the fractional parts
  of `N/n` and `N/p` (Saffari-Vaughan I and II as anchor, Graham-Kolesnik as
  method), from `import-fracparts.md` §6; and the twin tile as the sieved-set
  instance the existing hyperuniformity row records as absent, from
  `import-repulsive.md` §7. Both are outside the three rows this pass was sent
  with. The second one also needs the GLOSSARY decision first, since it would
  attach the tile to a convention whose word the corpus currently misuses.
- **`SEARCH-CONVENTIONS.md` §3**, the five "searches already run" rows drafted
  at `lit-dickman-variance.md` §6.2. Not applied; §3 is read by people and not
  by the check, so nothing depends on them being in place today.
- **`lit-dickman-variance.md` §6.3's amendment sentence** names "IMPORT-MAP row
  15", but the row 15 it quotes (Gorodetsky, `λ(s)`, `χ²/df`) is not the row 15
  written into the map here, which is the fractional-parts row. Whoever holds
  that note should say which row it means before the amendment is applied
  anywhere.
- **`TODO.md`** items Z5 and 4: their `Ledger:` lines need
  `Q-applied-0828-registries` added, and `research/QUESTIONS.md` needs
  regenerating with `node research/qc.js --index`. Both outside the fence.
- **`import-vc-nets.md` §6**, the `arXiv:2208.06442` bullet, may move to "PRIOR
  ART, CLEARED AT THE PAPER; NEIGHBOURHOOD ONE PAPER SHORT" with its title
  corrected to "multiples of the primes". A staging note, outside the fence.
- **Custody**, unchanged: `record-location-null.js` and
  `lit-kourbatov-shortfall.js` carry no OUTPUT banner, so the `b` figures now
  cited in `PRIOR-ART.md` and `SEARCH-CONVENTIONS.md` are the published ones,
  read at page image, and no in-house figure from those two notes was carried
  into a live file by this pass.

---

## 4. What would falsify this pass, and whether that check has run

- **A row was written into the map in the wrong words.** Falsified by comparing
  the three rows against `applied-0828-litimports.md` §5b. **RUN**, once, by the
  agent that made the edits; the cells match apart from the two status texts
  named in §1a, which come from the notes' own status paragraphs.
- **A table broke.** Falsified by counting unescaped column separators in every
  row touched. **RUN**: the three map rows carry ten columns, the three
  convention rows five, the REFUTED row five, the determinantal row three.
- **The `6/11` reading is wrong**, in which case row 15's struck THEOREM column
  and the `natal-cap-36` sentence both reverse. **PARTIALLY RUN**, inherited,
  one image of one scan, cross-checked against the paper's own `c/(c+2)`
  derivation.
- **The gate rejects something.** **NOT RUN**: `research/qc.js` is outside the
  fence for this pass. The known exposure is the ledger guard on TODO items Z5
  and 4, stated in §0.
- **The fence was breached.** **NOT RUN**: checking it means a git command, and
  the fence forbids one. Every write in this pass named one of the seven paths
  listed at the top.

---

*This document states current understanding at 2026-08-28. The six source notes
remain HELD; the items in §3 are drafts for whoever holds those documents.*