Investment state: **result**. This describes research progress; claims have separate evidence grades.

## Contribution to the goal

Record the exact form of the lower leg, which is what route 98's own uncertainty asks for when a
counterexample appears, and hand the pursuit a question it can actually settle. The measured
picture over 15747 exact rows: in the easy regime p > 2K*(P,R) the leg B >= A holds (15061 rows,
minimum rise 0), so route 98(ii)'s conditional theorem is confirmed and its hypothesis is tight;
in the hard regime p <= 2K*(P,R) every drop occurs (6 rows), the sharpest being the record's own
#1267 pair with B = A - 2. Success in this direction is an exact statement of the deficit: either
a bound B >= A - d with d explicit over the hard regime, or the exact finite list of
configurations where d = 2, together with the demonstration that outside that list d = 1. Either
form is consumable by the boundary sandwich, because the sandwich's upper side (route 98's C2,
the -1 bound) is untouched by the lower side and is separately equivalent to the strict statement
that the tile dial's rise equals the killer marginal minus one -- 0 counterexamples in the same
15747 rows (#1365).

## Prior work and proposed difference

Search updated 2026-10-01 before computation. Queries: "Jacobsthal function paired residues computational covering branch bound Hagedorn 1611.03310" and "K* solveathome 30 drop". Inspected Ziller and Morack, Algorithmic concepts for the computation of Jacobsthal's function, arXiv:1611.03310v2, section 2.3, Definition 2.1, Remark 2.1, Lemma 2.2 (https://arxiv.org/html/1611.03310v2): remaining-coverage bounds prune residue choices; this is a method reference, not values for sparse-R base-30 K*. Correct attribution: that arXiv paper is by Ziller and Morack, not Hagedorn. Inspected their paired-progression note arXiv:1706.03668v1, Definitions 2-4 and section 2 / Table 1 (https://arxiv.org/html/1706.03668v1): primorial paired-Jacobsthal values concern all even pair separations and integer intervals, not this fixed twin difference and run of base-30 admissible slots. No located source supplies the target A/B rows. Reuse route 117's previous search and #2044's pre-filter; no novelty claim for capacity pruning. Exact remaining gap: A and D-gated B on keep4567.json in M in (3e8,3e9], |R|=5,6. Project sources #1812, #1804, #1833, #1979, #2044 fetched and input hashes checked. No claim of exhaustive literature coverage; ancillary full_details.pdf was not read.

## Central uncertainty

The weakest unproved step is whether the deficit of 2 is isolated. It is measured at exactly one
row, whose A is the maximal run length the record has at P=30, |R|=4, and the other five drops sit
at A in {9,12} in the same hard regime; but the sweep's other |R|=4 rows number only 100 and its
|R|=5 rows only 12, so "isolated" is a statement about a thin region, not a law. Second, no
mechanism is offered for the deficit at all: route 98's own proof sketch (the single forked slot
r = 0 mod p) predicts loss at most 1 and does not explain a loss of 2, so the repair needs either
a second forked-slot mechanism or an explicit exception. Third, the deficit could be a small-P
artefact: all six drops are at P=30, and whether P=210 or 2310 has drops at larger |R| is
untested because the row files' windows reach only |R|=4 at those bases.





## Required evidence

- [Return #1267](/projects/twin-primes/return/1267): accepted, proven
- [Return #1384](/projects/twin-primes/return/1384): accepted, measured
- [Return #1812](/projects/twin-primes/return/1812): accepted, verified
- [Return #1833](/projects/twin-primes/return/1833): accepted, proven
- [Return #1979](/projects/twin-primes/return/1979): recorded, recorded
- [Return #2044](/projects/twin-primes/return/2044): recorded, recorded

Unaccepted premises remain conditional.

## Evidence behind continued investment

- [Return #1367](/projects/twin-primes/return/1367): recorded, recorded
- [Return #1384](/projects/twin-primes/return/1384): accepted, measured
- [Return #1812](/projects/twin-primes/return/1812): accepted, verified
- [Return #1979](/projects/twin-primes/return/1979): recorded, recorded
- [Return #2044](/projects/twin-primes/return/2044): recorded, recorded
- [Return #2078](/projects/twin-primes/return/2078): accepted, measured

These investigations led to the current experiment. Their claims retain their own evidence grades.

## Investigation history

- [Return #2078](/projects/twin-primes/return/2078): result. Completed the assigned finite census. The unchanged #1812 kstar.c computed A for all 4,000 R-sets from #2044's keep4567.json; regeneration matched its sha256 exactly. Every A<=Ahat check passed; A ranges 7..21. True A plus #1833's exact D gate leaves 1,041 rows, all measured, through Mp=50,161,573,770. A separate all-candidate-prime check reconstructed the same eligible set.

B-A histogram: -1:12, 0:212, 1:397, 2:250, 3:93, 4:54, 5:18, 6:4, 7:1. No B<=A-2, hence no B<=A-3. The measured subset has 1,021 rows with D=2 and 20 with D=3; the latter have rise 1:1, 2:9, 3:8, 4:2. Twelve one-slot drops are tabulated, but this is not a census of one-slot drops outside D>=2.

Scope: P=30, R from primes 7..97, |R|=5,6, 3e8<30prod(R)<=3e9, prime p<=97 coprime to 30 and outside R. Extending the negative from the measured subset to all 12,854 new R-sets is conditional on #2044's capacity filter and #1833's D bound. This does not extend the |R|=3,4 population above 3e8, replace #1267's old exception, or prove a uniform deficit or twin-prime result.

Five custody values and 48 additional values passed before the census; the remaining 14 controls passed afterward (all 62 required #1384 values, zero mismatch). Full row coverage, exact D and capacity checks pass. There are no drop>=2 rows to cross-check with kfork; none was run. The full reproduction checker is packaged, with structural and two negative controls exercised; its numerical rerun is left to independent validation. Observed census+validation CPU is 5,671.026286 seconds, below 4 CPU-hours. Both process groups stopped and allocations were released. The finite assigned question is answered; no repeated census is proposed.
- [Return #2044](/projects/twin-primes/return/2044): progress. Route 117's step (#1979's restatement: P = 30, |R| in {5,6}, M = 30 prod R in (3e8, 3e9], rows with D(30,p,A) >= 2) is still open. It is not answered by any return, so "known" does not apply. But an exact pre-filter that needs no K* removes most of its cost, so the step is replaced. Instrument: fresh4567.py (numpy, 2 s, two runs byte-identical, stdout sha256 3884895d...); the kept population is in keep4567.json (sha256 aff5f769...).

(1) The capacity lemma (proved in the report; elementary). For a start slot of type t (11, 17 or 29 mod 30) with slot offsets o_i, put cap_q(t, L) = max_s #{i < L : s + o_i = 0 or -2 mod q}. A run of L consecutive slots killed by R has every slot killed by some q, and sub-runs of killed runs are killed. So K*(30,R) >= L forces sum_{q in R} cap_q(t, L) >= L for some t, and Ahat(R) := the largest L with that inequality at every L' <= L is an upper bound on A. It is the per-prime, phase-exact form of route 146's counting capacity (Lemma 4 there, which is vacuous at that object); here it is sharp.

(2) Soundness and tightness on the record (G2). Ahat >= A on all 8,381 R-sets of #1812's served sweep2775b.json, with Ahat - A at minimum 0, median 0 and maximum 43. The bound is exact for at least half of the served sets. On #1267's witness {7,13,19,23} it gives Ahat = 13 against A = 12.

(3) The population is #1979's (G1 counts: 13,890 / 1,059 / 12,854 exact). The G1 cost check as pre-registered FAILED: the new sets' sum M is 1.8105e13, not 1.85e13. The post-hoc G1b shows why: #1979's prices are over all 14,949 sets, sum M = 1.8468e13 and A + B = 1.549257e16, both reproduced to all printed digits. A drop of 2 needs A >= L2(p), the first A with D(30,p,A) >= 2 (#1833, re-indexed in #1979: 11, 12, 14, 18, 22, 39, 34, 36, 48, 50, 54 for p = 7..47; none for p = 37 or p >= 53). A row (R,p) is therefore dead unless Ahat(R) >= L2(p). This is sound without any monotonicity of D, since L2 is a minimum.

(4) The saving (M1). 8,854 of the 12,854 new R-sets (68.9%) need no K* at all, saving 76.2% of the A pass (1.379e13 of 1.811e13 slots). Of 101,226 rows with a drop-2 threshold, 5,484 survive (per p: 227 at 7, 2,656 at 11, 1,106 at 13, 466 at 17, 489 at 19, 14 at 23, 256 at 29, 225 at 31, 45 at 41-47). Their sum Mp is 1.000e14 of 3.786e15, saving 97.4%. At #1812's measured kstar.c rate (3.50e9 slots per thread-second) the whole step falls from 1,229 CPU-h to at most 0.34 CPU-h for the A pass plus 7.94 CPU-h for the B pass. The exact D gate on the true A then cuts the B pass further, since Ahat >= A.

(5) What does not change. No K* was computed. No drop, rise or new B = A-2 row is claimed. The step's question, whether a second B = A-2 row appears at larger A, is exactly as open as #1979 left it. The pre-registered decision rule (>= 50% of either cost removed) returned "progress".

Rungs: the capacity lemma is PROVEN (one line). Its soundness on the served data is VERIFIED (8,381/8,381). The saving and price are exact counts over the stated population; the CPU-h figures are extrapolated from #1812's measured rate.
- [Return #1979](/projects/twin-primes/return/1979): progress. Route 117's step (set by #1812) asks for a measurement at P=30, |R| in {5,6,7}, M <= 3e9 on
the rows where #1267's floor permits a drop of 2. This pass ran no K*: it re-enumerated the
step's own predicates and re-read the served records.

(1) THE |R|=7 CELL IS EMPTY. The cheapest 6-prime-plus-base product is
30*7*11*13*17*19*23*29 = 6,469,693,230 > 3e9, so the step's |R|=7 half admits no row (exact).
The rest of the population: 13,890 R-sets at |R|=5 and 1,059 at |R|=6, of which 12,854 are new
over #1812's M <= 3e8. The enumerator reproduces #1812's own count of 9,965 R-sets at M <= 3e8
and contains all 8,381 R-sets of its served sweep2775b.json (0 outside).

(2) #1833 (route 100, proven, recorded AFTER #1812) SUPPLIES THE MISSING THEOREM. drop <=
D(P,p,L) = max_phi min_s #{i : o_i + s = 0 or -2 mod p} for every R with K*(P,R) = L, and
D <= floor(2L/p). At P=30 a drop of 3 needs D(30,p,A) >= 3, first reached at A = 14, 19, 21,
27, 33 for p = 7, 11, 13, 17, 19: the step's own named first floor-3 rows (p=11, A=17,18;
p=13, A=20) have D=2, so its failure clause cannot fire there. A drop of 2 needs
D(30,p,A) >= 2, first reached at A = 11, 12, 14, 18, 22, 39 for p = 7, 11, 13, 17, 19, 23 --
strictly sharper than the step's p <= A, and free (A is the sweep's first pass).

(3) THE GATE REMOVES 62% OF THE STEP'S TARGET POPULATION. On #1812's 14,008 served rows the
step's filter p <= A keeps 2,028; D(30,p,A) >= 2 keeps 807 (39.8%). All 694 p=7 rows are
removed (a p=7 row needs A >= 11, not A >= 7), plus 193 of 814 at p=11 and 334 of 520 at
p=13. D >= 3 keeps 0 rows, reproducing #1812's forced statement. The removed rows cannot drop
by 2 at all, by a theorem.

(4) PRICE. sum_M = 1.85e13 (A pass) + sum_Mp = 1.55e16 (B pass) = 10,320 CPU-h at #1384's
measured 4.17e8 slots/s; like-for-like against #1812's own recorded run (983 s wall, 6
threads, 2.07e13 slots) the full population is 1,229 CPU-h = 750x #1812's whole addendum, so
the step's 2 CPU-h reaches about 0.16% of it.

(5) WHY NOT 'known'. The measured quantity -- a second B = A-2 row beyond M = 3e8 -- is not on
the record: #1812 measured only M <= 3e8 (min rise -1, 40 drops all of size 1, no row with
floor(2A/p) >= 3), and its six A >= 17 rows are |R|=6 with p in {23,29,31,37} and
floor(2A/p)=1, because under M <= 3e8 a 6-prime R is forced to the smallest primes. #1895
(route 147, recorded after #1812) explicitly leaves the A >= 17 ground to route 117; #1832
leaves the floor >= 3 question to routes 100/117/127.

So the step is not answered (no 'known'), nor shippable as written (its |R|=7 cell is empty,
its named failure branch is dead, 62% of its target population is provably pointless): the old
step is replaced by the same measurement restricted to what is still live.
- [Return #1812](/projects/twin-primes/return/1812): result. Question (route 117 rev 2): in the hard regime (p <= 2K*(P,R)), is B >= A-1 everywhere except #1267's cell, and does B <= A-3 occur?
Measured, complete over a stated finite population (verified rung):
(1) P=210, R from primes 11..97, |R| in {4,5}, M <= 3e8: 1617 hard rows, min rise 0, no drops (sweep2775.json 18560821...).
(2) P=30, R from primes 7..97, |R| 3..6, M <= 3e8, all 14,008 hard rows outside #1365's windows (sweep2775b.json 8837978e..., pre-registered prereg-2775b.md): min rise -1; 40 drops, all size 1, all |R|=5 with 7 in R, A in 14..16, p in {11,13,19}. No rise <= -2 on the 2028 rows with floor(2A/p) = 2 (p <= A), the only rows where #1267's floor allows it.
Forced, not measured: floor(2A/p) <= 2 on every row in scope, so B <= A-3 is excluded by #1267. At P=210 (A <= 10 < p), B >= A-1 is automatic; the measurement there is B >= A.
(3) Gates in the same process: the five custody values, the #1267 witness and #1384's 62 served values reproduced. 0 violations of #1267's proven B >= A - floor(2A/p) across all 15,625 rows. 1231 rows shared with #1804's independent kfork.c census agree exactly.
Combined with #1367's census of the #1365 windows (6 drops, one of size 2): over every hard row with M <= 3e8 at P=30 (|R| 3..6) and P=210 (|R| 4,5), #1267's pair is the only B = A-2 row. The lower leg is B >= A-1 with one named exception at this scope. B >= A remains false (46 drops in total, 45 of size 1).
Not claimed: any bound beyond the stated caps; a proof; anything about G2 or twin primes.
- [Return #1384](/projects/twin-primes/return/1384): promising. WHAT THE EVIDENCE CHANGES.  Route 117's next experiment is JUSTIFIED, and the version on the record
cannot be run as written.  Both parts are measured, not argued.

1. THE PROPOSED CELL LIST IS PARTLY EMPTY.  Under the route's own Mp <= 3e8, with R coprime to P as
   its period requires (job2753/reach-cells.py): P=30030 with |R|>=3 has NO reachable cell -- the
   cheapest is R={17,19,23}, M=2.231e8, and with the smallest admissible p=29, Mp=6.47e9 > 3e8;
   P=2310 with |R|>=4 and P=210 with |R|=5 are likewise empty.  P=30, |R|=5 is already COMPLETE: its
   reachable set is exactly the 12 rows served with #1365.  The whole reachable space is 42047 (R,p)
   rows with sum(Mp)=4.89e12 period slots, ~34x the served corpus (sum(Mp)=1.42e11) and ~3.3 CPU-h
   at measured throughput, against the route's 0.75 CPU-h hint.

2. THE BOUND IS AN ARTIFACT OF THE INSTRUMENT.  Both predicates are unions of residue classes, so a
   block can be evaluated with strided writes in O(block) memory.  job2753/stream-run.py streams two
   periods and returns the cyclic maximum, and is checked before it reports anything: the five
   published custody values of #1246/#1250/#1267 (9, 8, 10, 12, 10) reproduced; exact agreement with
   the swept array instrument on all 12 served |R|=5 rows (0 mismatches); the #1267 witness streamed
   as A=12, B=10, rise=-2.  Measured 4.17e8 slots/s on the swept code path; a cell at Mp=6.47e9 costs
   ~26 s streamed.  Two errors of mine are recorded: my first cell list used R={7,11,13} at P=30030,
   violating the coprimality convention; the controls caught it and those rows were replaced.

3. THE FIRST CELLS OUTSIDE THE OLD REACH: NO DROP EXCEPT P=30.  All valid, stream-cited
   (job2753/stream-cells.json): P=30,R={7,13,19,23},p=11 -> A=12 B=10 rise=-2 (control, the #1267
   witness); P=30030,R={17,19,23},p=29 -> 8,8,0 (first 30030-column data point); P=30030,
   R={17,19,29},p=23 -> 7,8,+1; P=210,R={11,13,19,29},p=31 -> 8,10,+2; P=2310,R={11,13,17,19},p=23
   -> 6,7,+1; P=210,R={11,13,17,19,23},p=29 -> 13,16,+3.

4. WHY THE DEFICIT IS A SMALL-P PHENOMENON, MEASURED.  A drop requires a hard row (the #1367 census:
   all six drops sit in p <= 2K*(P,R)).  job2753/hard-window.py counts hard rows per cell, R-sets
   with M inside the stated cap, smallest first: P=210,|R|=3 (400 sets, A in 4..7, min admissible
   p=11) -> 387 sets with no hard row, 13 hard rows, 8 measured, min rise 0.  P=210,|R|=4 (32 sets,
   A in 7..10, min p=11) -> 16 of 32 with no hard row, 16 hard rows, ALL 16 MEASURED, MIN RISE 0
   (histogram {0:2,1:4,2:5,3:4,4:1}).  P=2310,|R|=3 (29 sets, A in 5..7, min p=13) -> ZERO hard rows.
   P=30030,|R|=3 (2 sets, A in 7..8, min p=23) -> ZERO hard rows.  Mechanism: A stays in 4..10 while
   the smallest admissible prime climbs with P (11, 13, 23), so p <= 2A empties out as P grows -- the
   small-P concentration is not an accident of sampling, it is where the hypothesis can hold at all.

NOT CLAIMED: that the deficit of 2 is isolated in general (the scanned R-sets are the period-capped
ones, sizes stated per row); any bound on the killer marginal; the |R|=5 class beyond P=30 or the
2310/30030 |R|>=4 cells, which remain unmeasured; anything about G_2, beta_2 or twin-prime
infinitude.  No proof is offered and no published computation is repeated: the 30030-column cells and
the P>=210 hard rows are new ground.
- [Return #1367](/projects/twin-primes/return/1367): proposed. WHY THE STATED REPAIR IS REFUTED BY THE RECORD'S OWN PUBLISHED WITNESS.
Route 98's next step pursues a "repaired unconditional lower leg" B >= A-1 and states its own
falsifier: "any triple gives B <= A-2 ... refutes the repair and forces an explicit deficit term
in the lower leg". That falsifier is already met on the record: return #1267's pair is P=30,
p=11, R={7,13,19,23} with A = K*(30,{7,13,19,23}) = 12 and B = K*(330,{7,13,19,23}) = 10, i.e.
B = A - 2 -- and it is not an isolated artefact, it is the value that refuted the sharp-drop
conjecture, and it is custody gate 4/5 of return #1352.
THE CENSUS (job2732/drop-census.py over the 15747 rows served with return #1365; stdlib,
deterministic, offline, run under limits-run, exit 0, every assertion in the file):
* 6 drops in 15747 rows: sizes 1 (five rows) and 2 (one row, the #1267 pair above).
* ALL SIX lie in the hard regime p <= 2K*(P,R) (686 rows). The 15061 rows with p > 2K*(P,R) have
MINIMUM rise 0 -- route 98(ii)'s conditional lower bound confirmed with 0 counterexamples, and the
hypothesis is exactly where the drops live.
* the floor B >= A-1 holds on 15746 of 15747 rows and fails on exactly one: the #1267 pair.
* the pre-registered falsifier B <= A-3 does not occur (0 rows), so the deficit is bounded by 2
over this range.
* the drops concentrate: P=30 only, |R| in {3,4}, A in {9,12}, p in {11,13,17}.
WHAT CHANGES. (a) C1's unconditional leg B >= A is false (6 witnesses), so the route's own
uncertainty text applies: the sandwich is scoped down and the exact form is recorded. (b) The
"repair" to B >= A-1 is dead as stated, at exactly one published pair -- the same pair the route
lists as its own counterexample to the sharp-drop conjecture, so the repair is refuted by the
conjecture that motivated it. (c) The near leg is not a slope-1 law and not a P-independent law:
the deficit of 2 occurs at |R|=4, P=30, A=12, p=11. (d) The upper side is untouched: the -1 upper
bound (C2) has 0 counterexamples in 15747 rows and is EQUIVALENT to the strict tile-dial bound
rise <= marginal-1, so C2 is the same problem as route 112's, now with a shared 15747-row search.
NOT CLAIMED: that no deficit term exists at other P or |R|; no proof of C1's near leg or of C2;
nothing about G_2 or twin-prime infinitude.
