Skip to content

Where characteristic 2 starts to pay: a T² threshold for root-unit decoders (E6 ties, E7 first gain found; no κ claim) - #2125

Merged
Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
DreamingOfClouds:research/char2-threshold
Oct 16, 2026
Merged

Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
DreamingOfClouds:research/char2-threshold

Conversation

@DreamingOfClouds

Copy link
Copy Markdown
Contributor

Where characteristic 2 starts to pay: a T² threshold for root-unit decoders (E6 ties, E7 first gain found; no κ claim)

Package research/char2-threshold. Model, cost formula and lemmas are those of the merged #1590 (exact decoders, Lemma B, ERR1) and of #2111 (three-total decoders); this PR adds the explanation of where an F2 gain over v first appears.

The law (argued). For root units the dense F2 cost of the full orthogonal pattern is about T × (block rank), the block rank is about T, so a gain over v needs more than about T² lines.

Table (CHECKED, exact; cost of an admissible pairing for the exact decoder M = Gram): D4/D5/D6 tie (cost = v = 12/20/30); E6 ties exactly at 36 = 6² (stored basis and B; B0 costs 37 in the first-independent basis); A4–A6 above v (B0 upper bounds 16/25/36; annealing found 16/21/34); E7 is the first gain found (49 < 63, T = 7; also checked in #1590, G3.E7-dyadic); E8 (72 < 120) is known (#1590 ERR1). The E6 B0 value 37 is checked in #1590 (G4.E6). F4 ties at 24 (recorded).

Fano Levi graph (CHECKED, exact): it realises over Q at T = 4; the exact minimum F2 cost of the stored realisations (rank(PQᵀ) = T, all T blocks nonempty) is 9 (T = 4) and 13 (T = 5) against v = 7: no gain at that size.

Realisability (RECORDED, numerical): the Sp4 doily (15 points) realises at T = 4 and not at T = 3; the 63-point Sp6 realises at T = 7, and Levenberg–Marquardt found no realisation at T = 6 or 5. No exact rational realisation is shipped, so these are not CHECKED.

Scope. The searches (annealing, cliques, lattice pools) are not exhaustive and are recorded, not re-run; the table costs are costs of specific B (upper bounds on the F2 minimum). There is no claim that T ≤ 6 can never beat v; the law is an explanation plus evidence. No κ claim, no window closed. Context: #606, #1698, #1830.

Verify. bash research/char2-threshold/verify.sh /fresh/dir (Python standard library and cc; seconds; step 0 checks MANIFEST.sha256 and SOURCES.json and stops hard on mismatch; 10 PASS lines including step 0).

Prepared by DreamingOfClouds with Claude Opus 5.5 (Anthropic), via Claude Code; Apache-2.0.
🤖 Generated with Claude Code

…coders (E6 ties, E7 first gain found; no κ claim)

Package research/char2-threshold. Exact F2 costs of the full-orthogonal-pattern pairings of root units against T^2 (D4-D6 and E6 tie, E7 49 < 63
first gain found, E8 72 < 120), exact minimum cost of a Fano Levi-graph realisation over Q (9 at T = 4, 13 at T = 5 with rank(PQ^T) = 5, against v = 7),
numerical realisability of Sp4/Sp6 recorded. Law is an explanation plus evidence; no claim that T <= 6 cannot beat v. No kappa claim.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@Mister-Meeseeks
Mister-Meeseeks merged commit ee997d5 into CrocSwap:main Oct 16, 2026
47 of 62 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants