An intermediate conjecture between Goldbach and Dubner: every even number is the sum of a prime and a twin prime

Tushar Pandey

REVIEW 2 major objections 5 minor 12 references

An intermediate conjecture between Goldbach and Dubner: every even number is the sum of a prime and a twin prime

T0 review · 2 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Every even number up to 10^12 is the sum of a prime and a twin prime — a statement that, if true for all evens, would prove both Goldbach's and the twin-prime conjecture.

desk verdict A sound elementary rigidity theorem and a plausible but single-sieve verification to 10^12; the math is fine, the empirical headline needs an independent sieve or at least a documented boundary margin. read the letter →

arxiv 2608.02381 v1 pith:4W6JIC5N submitted 2026-08-03 math.NT

Tushar Pandey This is my paper · ORCID

classification math.NT MSC 11P3211A4111Y11

keywords twinprimesGoldbachconjectureDubner'sprimeplusorientationrigidityleastwitnessexhaustiveverificationHardy-Littlewoodheuristic

topics Goldbach's Conjecture The Twin Prime Conjecture

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper isolates a single statement (S): every even n ≥ 6 is the sum of a prime and a twin prime (a prime belonging to a pair differing by 2). It shows that (S), if true, would imply both the Goldbach conjecture and the twin prime conjecture, making it the weakest known statement with that double consequence. The paper proves a rigidity theorem: for even n not divisible by 3, all twin members in Goldbach partitions of n have the same orientation — lower or upper — except for trivial exceptions involving 3 and 5. It then verifies (S) exhaustively for all even n up to 10^12, with the largest least twin witness being 14,549, and matches Hardy–Littlewood predictions for the density of twin-touching partitions. Along the way it re-derives Dubner's exception list and extends the verification of Dubner's conjecture to 10^11.

What carries the argument

The key structural input is Theorem 2, the orientation-rigidity theorem: a congruence argument modulo 3 shows that if n ≡ 1 (mod 3), any twin member t in a Goldbach partition n = p + q must have t+2 prime; if n ≡ 2 (mod 3), t must have t−2 prime — with the lone exceptions t = 5 and q = 3, or t = 3 in the second case. This theorem forces the orientation of twin witnesses and underlies the heuristic singular series and the mod-6 triple structure that explains Dubner's exception list. The computational engine is a bit-packed Boolean sieve with segmented shift-and-OR accumulation, a standard technique for Goldbach-style verifications, executed in two independent kernels for the run to 10^11 and

What would settle it

A single even n ≤ 10^12 with no representation n = p + t, or with t_min(n) > 14,549, would refute the main computational claim. Likewise, a single even n not divisible by 3 with a Goldbach partition containing a twin member of the opposite orientation (outside the listed 3-5 exceptions) would refute Theorem 2. For the heuristic, measuring the maximum of t_min over the evens up to 10^13 or 10^14 and comparing it with (4/c_min)(log N)^2 (log log N)^2 would test the predicted exponent.

Watch

Extended reading notes

Core claim

Conjecture (S) — every even n ≥ 6 can be written as p + t with p prime and t a twin member — sits strictly between the Goldbach and Dubner conjectures. The paper's central discovery is twofold. First, (S) is a single elementary statement that implies both the Goldbach conjecture and the twin prime conjecture (Theorem 1). Second, any proof of (S) would have to contend with a mod-3 orientation rigidity: for even n not divisible by 3, every twin member appearing in a Goldbach partition of n is a lower member (t+2 prime) when n ≡ 1 mod 3, and an upper member (t−2 prime) when n ≡ 2 mod 3, with the only exceptions being the primes 3 and 5. The paper verifies (S) exhaustively to 10^12, computing th

Load-bearing premise

The exhaustive computational verification to 10^12 rests on the bit-packed sieve correctly identifying all primes up to that bound; a systematic bug in the sieve (the paper notes that the double cover tests the scan kernels, not the sieve itself) would invalidate the 'zero exceptions' claim and the t_min records.

Editorial extensions

If this is right

If (S) is true, then both the Goldbach conjecture and the twin prime conjecture follow immediately; no other single statement at this level is known to imply both.
The mod-3 rigidity implies that any proof of (S) must split into cases n ≡ 0, 1, 2 (mod 3), with the n ≡ 0 case allowing both orientations.
The exhaustive verification shows t_min(n) ≤ 14,549 for all even n ≤ 10^12, far below log^3 n, suggesting t_min grows very slowly; Proposition 1 shows a bound t_min(n) ≪ log^A n for any A would yield a power lower bound π_T(z) ≫ z^{1/A} on twin primes, a dramatic strengthening of the twin prime conjecture.
The re-verification of Dubner's conjecture to 10^11 extends the known range by a factor of five and reproduces the 33-term exception list, explained by the mod-6 triple structure.
The Hardy–Littlewood heuristic predicts max t_min(n) ≍ (log N)^2 (log log N)^2, which is compatible with the observed records at each decade.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the heuristic for the maximum holds, then at 10^13 or 10^14 the record should grow by only a few thousand; a jump beyond the (log N)^2 (log log N)^2 shape would discredit the model and possibly the underlying heuristic for (S).
Proposition 1 suggests a concrete route to a power bound on twin primes: try to prove t_min(n) ≪ log^3 n (or even a larger exponent) using methods that are weaker than full TPC. Even an exponent like A = 10 would give π_T(z) ≫ z^{0.1}, which is currently unknown.
The mod-6 triple structure implies that Dubner exceptions come in full triples; searching for new exceptions beyond 10^12 could focus on centres 6m where the ordered pair count L is zero, which is checkable efficiently.
The orientation-rigidity theorem may offer a new sieve-theoretic obstruction: a proof of (S) would need to handle the parity problem while respecting the mod-3 orientation, perhaps suggesting that a level-of-distribution approach could be attempted for the triple pattern (t, t+2, n−t).

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The two things worth knowing: this paper proves a real, small theorem about the mod-3 orientation of twin members in Goldbach partitions, and it reports an exhaustive verification of (S) to 10^12 that is probably correct but not independently audited. The conjecture itself is not new—properly credited to [7]—and the hierarchy observation that (S) implies both GC and TPC is elementary but useful.

The proven math checks out. Theorem 2 is genuinely clean: for n not divisible by 3, all twin members in Goldbach partitions share an orientation, up to exceptions involving 3 and 5. The mod-3 forcing argument is correct. Theorem 1's counting is correct. Proposition 1 is a nice calibration: a log^A n bound on the least witness gives a power-of-z lower bound for twin primes. I also checked the local factors in Corollary 1; they're right. The paper gets credit for laying out the heuristic layer as heuristic, with the free constant explicitly uncalibrated.

The soft spot is exactly where the reader and the stress test put it: the 10^12 verification rests on a single segmented sieve. The paper candidly says the 10^11 double-cover tests the scans, not the sieve, and the 10^12 extension is one S-only pass. That's an honest disclosure, but it means the zero-exception claim and the t_min records inherit any systematic misclassification in that one code path. The stress test also raises a concrete gap in the write-up: for n near the start of a block, n-t can fall below the block's lower bound, and the paper never explains how the segmented sieve supplies primality for that boundary margin. That's not necessarily a computational error—there are standard ways to handle it—but as written the implementation detail is missing, and it matters for the headline claim.

The heuristic "out-of-sample confirmation" is another minor weakness: the model was supposedly committed before the 10^12 run, but there's no timestamped archive of the model to verify that. The paper's own caution about one decade being one data point is appropriately modest. These are not fatal flaws; they're auditability gaps in the empirical part, not in the theorems.

Who gets value from this? Number theorists working on Goldbach-type problems and anyone doing large-scale elementary sieve verification. The computational methodology discussion—single pass, shared sieve, boundary handling—is a useful case study for people running similar scans. It deserves a serious referee: send it to a computational number theory venue, and the referee should ask for either an independent sieve run or an explicit boundary-margin description plus a small randomized cross-check. I'd take the theorems as established and treat the 10^12 verification as plausible but not yet bulletproof.

Referee Report

2 major / 5 minor

Summary. The paper studies the statement (S): every even n >= 6 is a sum of a prime and a twin member (a prime t for which t-2 or t+2 is prime). The authors note that (S) is intermediate between the Goldbach conjecture and Dubner's conjecture, and that it implies both Goldbach's conjecture and the twin prime conjecture (Theorem 1). They prove a modular orientation rigidity theorem (Theorem 2): for n not divisible by 3, all twin members in Goldbach partitions of n have one orientation (lower or upper) determined by n mod 3, with degenerate exceptions confined to the primes 3 and 5. They also prove Proposition 1, showing that a hypothetical bound t_min(n) <= C log^A n would imply the power lower bound pi_T(z) >> z^{1/A} for twin members. The main empirical contribution is the claimed exhaustive verification of (S) for all even n <= 10^12, with maximum least twin witness 14,549 at n = 571,714,791,706, along with re-verification of Dubner's conjecture to 10^11 and a heuristic analysis of t_min. Theoretical results are checked and appear correct; the computational verification is the load-bearing empirical claim.

Significance. If the computational verification is correct, the paper provides a substantial numerical result for a natural intermediate conjecture, extends verified ranges for related problems, and identifies a clean rigidity constraint (Theorem 2) that any proof of (S) must respect. The theoretical derivations—Theorem 1, Theorem 2, and Proposition 1—are elementary, correct, and independent of the heuristics; their proofs are verifiable and constitute a real contribution. The paper is also commendably explicit that the t_min model is heuristic and that the verification has limitations, including the statement that the 10^11 double-cover tests the scans and not the sieve. The significance of the paper hinges on the reliability of the exhaustive sieve computation, which is not independently certified.

major comments (2)

[§5, segmented shift-and-OR description] The central empirical claim—zero exceptions for all even n <= 10^12 and t_min maximum 14,549—rests entirely on the correctness of the monolithic/segmented Boolean sieve and the twin-member table. As the paper itself concedes, the 10^11 double-cover 'tests the scans and not the sieve,' and the 10^12 extension is a single S-only pass. A systematic sieve error, for example a composite misclassified as prime, would cause n to be marked as covered by a false witness, so no candidate exception would be reported and t_min would be underestimated. The independent Miller–Rabin re-verification covers only the record witnesses, not the millions of primality decisions for n-t needed for the zero-exception claim. This is load-bearing for the paper's headline result. Please either provide an independent verification path (e.g., a second sieve implementation or a comparison against a different primalit
[§5, 'Both kernels cap the twin witnesses'] The description of the segmented shift-and-OR method does not state how primality of n-t is obtained when n is near the beginning of a segment and t can be as large as 10^7, so n-t lies below the segment's starting point. If the segmented sieve does not explicitly include a lower margin covering this range, then even numbers at the start of each block could be silently missed, defeating the candidate-exception mechanism. The paper says the recorded intervals tile [0,10^12] contiguously, but that does not address the boundary margin. Please describe the sieve layout (including the extended range below the block, or any other mechanism) and, if necessary, verify that block-start residue classes are covered.

minor comments (5)

[Abstract and §5] The abstract's unqualified 'We verify (S) exhaustively for all even n <= 10^12' should be harmonized with the caveats in §5 (single implementation, no independent primality check for the bulk of the range). As currently written, the abstract overstates the epistemic status of the computational claim.
[§2] The empirical check of Theorem 2 over only 1500 consecutive even numbers at two starting points is a sanity check, not evidence for the exhaustive claim, and should be labeled as such. The text's 'exactly 0' is likely to be misread as a general verification.
[§5, record witnesses] The phrase 'the three record witnesses below' is ambiguous because the table lists six records. Please clarify which witnesses were independently re-verified.
[§5, heuristic] The heuristic derivation of t_min(n) is clearly labeled as heuristic, which is good. However, the claim that the model was 'committed before the 10^12 run' cannot be checked from the manuscript; if this is important, consider adding a date or version note to the archived scripts.
[§4] The notation 'S(n)' for the combined singular series and '(S)' for the conjecture is potentially confusing; consider a different symbol for the conjecture, e.g., Conjecture G.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the theorems are proved in-text, the computational benchmark is external, and the heuristic layer is explicitly out-of-sample.

full rationale

The paper's central derivations are self-contained and do not reduce to their inputs. Theorem 2 is a modular-arithmetic proof from the definitions of twin-member orientation; it uses no fitted parameter and no prior result of the authors. Theorem 1 and Proposition 1 are counting/implication arguments proved in the text. The exhaustive verification to 10^12 is an external computational benchmark, not a consequence of the conjectures; the paper explicitly flags its limitations in §5 ('that agreement tests the scans and not the sieve', 'the extension to 10^12 is a single S-only pass'), which is a reliability caveat about the sieve, not a circular derivation. The §4–5 Hardy–Littlewood heuristic is labeled heuristic ('none of it is proved, and it is set down only to calibrate the question'), and the claimed out-of-sample confirmation of the extremal mechanism ('committed before the 10^12 run ... out-of-sample confirmation rather than a fit') is a test of an independent model on new data, not a fitted parameter renamed as a prediction. No equation is defined in terms of the target claim, no load-bearing result rests on a self-citation, and no uniqueness theorem is imported from the authors' prior work. Accordingly, no significant circularity is present.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The proved results (Thm 1, 2, Prop 1) are self-contained and rest only on standard facts; the verification results rest on sieve correctness; the heuristic layer rests on the Hardy-Littlewood conjectures. No fitted parameter is labeled a prediction; the heuristic constant is honestly uncalibrated. No new entities are postulated: 'twin member' is a definition of an existing prime set, 'lower/upper' is a labeling of the classical 6k±1 structure, and (S) is a conjecture about primes and twin primes only.

free parameters (1)

Uncalibrated multiplicative constant in the t_min heuristic
The model t_min(n) ≈ log n·(log log n)^2/S(n) and its extremal version (4/c_min)(log N)^2(log log N)^2 have an explicitly uncalibrated O(1) constant (§5: 'the O(1) constant uncalibrated'). It affects only the heuristic layer, which the paper labels as not proved.

assumptions (5)

domain assumption Hardy-Littlewood k-tuple heuristic with singular series Glow(n) + Gup(n)
Invoked in §4 to predict R_T(n), the c(n)/log n decay of twin-touching fractions, and the W(n,T) model for least witnesses; the paper labels these predictions heuristic. Corollary 1 establishes the series is bounded below uniformly.
domain assumption Correctness of the Boolean sieve and twin-member table in §5
Every empirical claim — zero exceptions to 10^12, t_min max 14,549, re-derivation of A007534 — depends on the sieve correctly identifying primes. The paper notes the 10^11 double cover 'tests the scans and not the sieve'.
standard math Chebyshev bound π(x) ≪ x/log x
Used in Proposition 1 (§5) to pass from π(N)π_T(z) ≥ N/4 to π_T(z) ≫ z^{1/A}.
domain assumption Dubner's verified range and A007534 are as re-derived by the single both-twin run to 10^11
§5: the both-twin run 'was carried out once rather than twice'; the A007534 re-derivation and Dubner re-verification to 10^11 inherit any bug in that single run. The (S)-only 10^12 leg does not depend on this by-product.
domain assumption Per-block recorded intervals tile [0, 10^12] contiguously
§5: exhaustiveness of the (S) check relies on the programmatic tiling check being itself correct ('checked to tile [0, 10^12] contiguously, without gap or overlap').

how reviews work

0 comments

Cite this review

Pith. "Pith review of An intermediate conjecture between Goldbach and Dubner: every even number is the sum of a prime and a twin prime." pith.science (2026). https://pith.science/paper/4W6JIC5N

@misc{pith2026260802381,
  author       = {Pith},
  title        = {Pith review of: An intermediate conjecture between Goldbach and Dubner: every even number is the sum of a prime and a twin prime},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4W6JIC5N}},
  note         = {Machine review of arXiv:2608.02381}
}

abstract

Call a prime $t$ a twin member if $t-2$ or $t+2$ is prime, and let $\mathcal{T}$ denote the set of twin members. We study the statement (S): every even $n \ge 6$ can be written as $n = p + t$ with $p$ prime and $t \in \mathcal{T}$. Statement (S) sits between the Goldbach conjecture and Dubner's conjecture (every even $n > 4208$ is a sum of two twin members), and we observe that (S) is a single elementary statement implying both the Goldbach conjecture and the twin prime conjecture. We prove an orientation-rigidity theorem: for $n \not\equiv 0 \pmod 3$, all twin members $t$ appearing in Goldbach partitions of $n$ (representations $n = p + q$ with $p, q$ prime) have the same orientation. According to $n \bmod 3$, either every such $t$ has $t + 2$ prime, or every such $t$ has $t - 2$ prime, with degenerate exceptions confined to the primes $3$ and $5$. We verify (S) exhaustively for all even $n \le 10^{12}$. In the course of the earlier run to $10^{11}$ we also re-derive the $33$ terms $\ge 6$ of the Dubner exception list A007534 and re-verify Dubner's conjecture; those two by-products are established for $n \le 10^{11}$ only, the extension to $10^{12}$ covering (S) alone. We compute the least twin witness $t_{\min}(n) = \min\{t \in \mathcal{T} : n - t \text{ prime}\}$, whose maximum is $14549$ (at $n = 571714791706$), below $\log^3 n$, and the density of twin-touching Goldbach partitions (those containing a twin member), which decays like $c(n)/\log n$ and matches the Hardy-Littlewood prediction for the constant $c(n)$: measured median values $4.93$-$5.01$ against a predicted median of $4.97$-$4.98$ (the mean is $8C_2 \approx 5.28$, with $C_2$ the twin-prime constant).

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

12 extracted references

[7]

S. M. Naser and K. F. U. Ahmed, On twin prime numbers, preprint (2025), available at https://www.researchgate.net/publication/392575845, accessed July 2026

arXiv 2025
[1]

Brun, Le crible d’ ´Eratosth` ene et le th´ eor` eme de Goldbach,Videnskapsselskapets Skrifter, Mat.-Naturv

V. Brun, Le crible d’ ´Eratosth` ene et le th´ eor` eme de Goldbach,Videnskapsselskapets Skrifter, Mat.-Naturv. Klasse, Kristiania(1920), no. 3, 36 pp

1920
[2]

Deshouillers, H

J.-M. Deshouillers, H. J. J. te Riele, and Y. Saouter, New experimental results concerning the Goldbach conjecture, inAlgorithmic Number Theory (ANTS-III), Lecture Notes in Comput. Sci.1423, Springer, Berlin, 1998, 204–215

1998
[3]

Dubner, Twin prime conjectures,J

H. Dubner, Twin prime conjectures,J. Recreational Math.30(1999–2000), no. 3. Scan available athttps://oeis.org/A007534/a007534.pdf

1999
[4]

Granville, J

A. Granville, J. van de Lune, and H. J. J. te Riele, Checking the Goldbach conjecture on a vector computer, inNumber Theory and Applications(R. A. Mollin, ed.), Kluwer, Dordrecht, 1989, 423–433

1989
[5]

G. H. Hardy and J. E. Littlewood, Some problems of ‘Partitio numerorum’ III: On the expression of a number as a sum of primes,Acta Math.44(1923), 1–70

1923
[6]

A. Kourbatov, Goldbach’s conjecture – and more: even numbers are sums of two primes (twins!), JavaScripter.net (1999–2011), formerly athttp://www.kourbatov.com/math /primes/evennumberisasumoftwoprimetwins.htm(no longer resolving), archived at https://web.archive.org/web/20260209043949/http://www.kourbatov.com/math/p rimes/evennumberisasumoftwoprimetwins.h...

arXiv 1999
[8]

OEIS Foundation Inc., Sequence A007534 (Positive even numbers that are not the sum of a pair of twin primes),https://oeis.org/A007534

Show all 12 references

[9]

Oliveira e Silva, S

T. Oliveira e Silva, S. Herzog, and S. Pardi, Empirical verification of the even Goldbach conjecture and computation of prime gaps up to 4·1018,Math. Comp.83(2014), 2033–2060

2014
[10]

Richstein, Verifying the Goldbach conjecture up to 4·10 14,Math

J. Richstein, Verifying the Goldbach conjecture up to 4·10 14,Math. Comp.70(2001), no. 236, 1745–1749

2001
[11]

M. K. Sinisalo, Checking the Goldbach conjecture up to 4·10 11,Math. Comp.61(1993), no. 204, 931–934

1993
[12]

Zwillinger, A Goldbach conjecture using twin primes,Math

D. Zwillinger, A Goldbach conjecture using twin primes,Math. Comp.33(1979), no. 147, 1071. 9

1979

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [7]

S. M. Naser and K. F. U. Ahmed, On twin prime numbers, preprint (2025), available at https://www.researchgate.net/publication/392575845, accessed July 2026

arXiv 2025

[2] [1]

Brun, Le crible d’ ´Eratosth` ene et le th´ eor` eme de Goldbach,Videnskapsselskapets Skrifter, Mat.-Naturv

V. Brun, Le crible d’ ´Eratosth` ene et le th´ eor` eme de Goldbach,Videnskapsselskapets Skrifter, Mat.-Naturv. Klasse, Kristiania(1920), no. 3, 36 pp

1920

[3] [2]

Deshouillers, H

J.-M. Deshouillers, H. J. J. te Riele, and Y. Saouter, New experimental results concerning the Goldbach conjecture, inAlgorithmic Number Theory (ANTS-III), Lecture Notes in Comput. Sci.1423, Springer, Berlin, 1998, 204–215

1998

[4] [3]

Dubner, Twin prime conjectures,J

H. Dubner, Twin prime conjectures,J. Recreational Math.30(1999–2000), no. 3. Scan available athttps://oeis.org/A007534/a007534.pdf

1999

[5] [4]

Granville, J

A. Granville, J. van de Lune, and H. J. J. te Riele, Checking the Goldbach conjecture on a vector computer, inNumber Theory and Applications(R. A. Mollin, ed.), Kluwer, Dordrecht, 1989, 423–433

1989

[6] [5]

G. H. Hardy and J. E. Littlewood, Some problems of ‘Partitio numerorum’ III: On the expression of a number as a sum of primes,Acta Math.44(1923), 1–70

1923

[7] [6]

A. Kourbatov, Goldbach’s conjecture – and more: even numbers are sums of two primes (twins!), JavaScripter.net (1999–2011), formerly athttp://www.kourbatov.com/math /primes/evennumberisasumoftwoprimetwins.htm(no longer resolving), archived at https://web.archive.org/web/20260209043949/http://www.kourbatov.com/math/p rimes/evennumberisasumoftwoprimetwins.h...

arXiv 1999

[8] [8]

OEIS Foundation Inc., Sequence A007534 (Positive even numbers that are not the sum of a pair of twin primes),https://oeis.org/A007534