Answers are grounded in this paper and cite their sources.
This record is version v0.2.0 of a non-peer-reviewed, extensively AI-assisted research project on the diagonal Ramsey number R(5,5). It is a substantial expansion and reorientation of the v0.1.0 release. The earlier version studied the reconfiguration geometry and low-defect neighbourhood of the public 656-class collection of known Ramsey(5,5,42) graphs. Version v0.2.0 retains that corrected catalogue-relative work as a companion, but the new primary manuscript is directed at the unresolved order-43 problem itself. The first new line is an extension-deletion coherence theory at orders 41-43. For a fixed 42-vertex Ramsey(5,5)-free graph G and a candidate neighbourhood row X for a new vertex, the manuscript identifies exactly which vertex-deleted cards accept the restriction of X: the accepting set is the intersection of the four-vertex supports of the extension defects. Because every defect support has size four, an invalid row can be locally valid on at most four deleted cards, whereas validity on five deleted cards forces a genuine global 43-vertex extension. The same framework yields a safe two-card gluing theorem, an exact three-card wedge criterion for deciding fixed-base 42-to-43 extension, and an equivalent order-41 rank-two compatibility formulation. These statements do not assume that the public 656 order-42 graphs form a complete catalogue. The second new line is a near-extremal phase reduction for a hypothetical 43-vertex Ramsey(5,5)-free graph. Reparameterising the McKay-Radziszowski excess identity by exact R(4,5) edge deficits gives a global deficit budget and a strict-balanced dichotomy. In the balanced phase, the manuscript proves that a hypothetical counterexample must be regular of degree 18, 20, 22, or 24. A complete computation over the 352,366 published classes in R(4,5,24), using two motif-counting implementations, gives a strictly negative order-43 quartic potential throughout the catalogue and excludes the balanced 18/24 complement pair. The surviving balanced phase is therefore reduced to the 20/22 complement pair. Rounds 32-40 progressively restrict that surviving phase. The low-nonedge graph in the balanced 20-regular branch is forced to have minimum degree at least five and at least 108 edges. The equality-five frontier is compressed to 19 coarse states; on the three tight q=1 states, the Round-39 signed-link analysis excludes all mu=5 roots and forces a mu=6 root. Round 40 adds an endpoint-capacity classification of all 74 x 843 = 62,382 order-18/order-24 pairs at e(A)=85: 61,939 pairs fail the capacity test and the remaining 443 equality pairs are excluded by a terminal repetition obstruction. Independently, all 74 supplied edge-85 order-18 cores are exhaustively tested over 74 x 2^18 = 19,398,656 candidate extension rows, with zero legal rows. This yields a hereditary induced-order-18 cap used to sharpen the equality-five row bounds. The three tight q=1 states are further reduced to 67 rooted mu=6 low-skeleton orbits and 6,480 necessary R(3,5,7)-side signatures. The corrected catalogue-reconfiguration companion is also included. Relative to the public 656-class order-42 collection, it records a legal single-edge quotient with 1,976 edges and 12 connected components, a threshold-10 one-vertex extension census with 102 labelled extensions and 91 ordinary isomorphism classes, a defect-two neutral C86 and deletion bridge, and the complete defect-at-most-one closure reachable from the known catalogue: 8,284 isomorphism classes and 33,154 quotient edges, containing no additional defect-zero class. The v0.1.0 reproducibility audit independently reran the principal finite counts and motivated proof-chain and presentation corrections incorporated in this companion version. Scope and non-claims: v0.2.0 does not prove R(5,5)=43, does not change the published bound 43 <= R(5,5) <= 46, does not assume that the 656 public order-42 graphs are exhaustive, and does not eliminate the full balanced 20/22 phase or all strict pivot classes. Several near-extremal statements depend on published complete Ramsey catalogues as imported inputs. AI-use and responsibility notice: ChatGPT and Kimi were used extensively in exploratory research, research-direction generation, manuscript drafting and editing, code generation and modification, consistency checking, organization of evidence, and release documentation. The AI systems are not authors. Chengjun Lyu and Chang Lyu are the named human authors, directed the research process and public release, and remain responsible for maintaining, correcting, revising, or withdrawing the artifact when warranted. Specialist independent validation of every mathematical, computational, and historical-priority claim remains incomplete. Evidence-status notice: the complete Round-20-to-40 source/data/checker archive referenced by the new manuscript was not among the files supplied to the present Zenodo assembly. The v0.2.0 record therefore includes the primary manuscript and computational supplement, together with the corrected and reproducible catalogue-reconfiguration package, but it should not be described as containing a fully independently rerunnable Round-40 proof archive. The precise boundary is documented in the accompanying EVIDENCE_STATUS file. Zenodo hosting and DOI registration provide archival dissemination and versioning; they do not constitute peer review, mathematical certification, historical-priority confirmation, or endorsement.
No takes yet. Share an insight, caveat, or question.
Lyu et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: