Skip to content

Exclude arbitrary fixed-chart relays at nine rank-seven BIT deadlines - #7750

Open
aliromman wants to merge 1 commit into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-065
Open

aliromman wants to merge 1 commit into
CrocSwap:mainfrom
aliromman:codex/consecutive-result-065

Conversation

@aliromman

@aliromman aliromman commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

Arbitrary finite paid relay networks on the retained charts cannot improve the nine original rank-seven BIT arrivals into chart 0. The original price 7^t is minimum for every 0<t<=1, with free retention, local XOR and access to all strictly earlier original stock. This permits new intermediate forms, branching, merging, cycles and cancellation. It extends PR #7741's direct-bundle result at these nine deadlines. No new kappa or relay guarantee at the other 5,733 arrivals is claimed.

Read the proof, tree/path receipt, verifier, exact source pins and credits.

Every improving network must have total integer rank at most six. Its relevant chart component has a spanning tree within that budget. If a dimension-six chart occurs, the whole tree is a strictly increasing nested chain from zero to six. Otherwise its dimensions are 0, 1, 2 and 4. These cases bound the support of every form available to the receiver, even after new relay combinations.

Complete necessary bound Evidence
Low-dimensional trees: support at most 37 Exact rooted-tree recurrence with arbitrary branching; an independent coarse cut bound also gives 42<48
Chains reaching dimension six: support at most 35 All 37,872 chains; exact source support unions
Separate path census Forward and reverse constructions agree on all 40,667 increasing paths
Required arrivals: distance at least 48 All 511 target-coset vectors in the unchanged predecessor certificate
Attained old price The original source and nested rank-seven transfer remain feasible

Thus every cheaper network has projected nonroot support at most 37, while every required target representative has weight at least 48. The old direct-only cap of 30 does fail on an actual relay chain; the new result charges and covers that changed information flow. The 37 bound is an upper bound for actual stock, not a claimed exact communication capacity.

Python 3.12, standard library only:

python3.12 -B scripts/research_package.py --package research/bit-nine-deadline-relay-floor

Both registered normal and optimized commands print:

PASS_BIT_NINE_DEADLINE_RELAY_FLOOR all relays: support<=37<48

Normal/optimized registered replay, committed bytes/source history/modes, export and sealed read-only Git-free serial replay pass. The complete retained PR7741 verifier runs first, including its source geometry, two paid-row representations, all attainments, donor supports, cosets and controls. New controls cover the direct-only cap, false strict separation, support unions and charged extra branches. Hosted CI did not certify this package. The plan job failed in the shared test_current_ci_section_matches_selected_record: the base guide names frontier6030 while the selected record requires make frontier7000-verify. The log reports 760 tests, one failure and 24 skipped; package replay was skipped. This package changes no shared CI file. Local replay remains the scientific evidence.

Ali Romman supplied the fixed quotient and coset proof and certificate. Nick Petalas supplied PR #7739's reduction and nine exceptions. PR #7723 supplied the fixed-library model, and Jerry Chen supplied the selected PR #7001 caller. The retained credit file preserves zxv, DreamingOfClouds, Cinna Davis and earlier native/root attribution. Their bytes and notices remain intact.

The model fixes charts, independent inputs, original stock, target rows and deadlines. New charts can change the obstruction: a proposed dimension-three hub with three dimension-four sources has rank bill 1+1+1+3=6, but no compatible source intersection, target reconstruction or physical caller is supplied. The other 5,733 arrivals need their own relay analysis. All five framework assumptions and the selected frontier remain unchanged. The selected cold replay, unrelated historical suites, Lean, Docker and physical relay execution were not run.

Base: canonical main e54523517d0f03e58352e647c403a5af3e525a57, revision 8. The complete eighteen-file predecessor package is retained at 33817d65c801232f1e07355f43ec6040a41e7c7e; its own eight source pins remain intact. No unmerged source ancestry is asserted. Geometry and source replay are shared code. Prepared by Ali Romman with substantial OpenAI Codex assistance in research, implementation, verification and writing; no independent human or formal-kernel acceptance. Apache-2.0. Canonical claim.

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.

1 participant