Repository navigation
Research: exact two-relation four-register pattern census and scoped obstruction - #1733
Conversation
|
Batch review at Merged to Reviewed scope: Complete four-register binary idempotent-pattern census and a fixed two-relation, fifteen-address ansatz are scoped separately. Common-order exclusion uses the reduced triangular formal algebra and rank subadditivity; the 120 exceptional prefixes exclude only that specified schedule, not every two-relation rational word. The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions. Batch review, attribution and evidence. Thank you for the contribution. |
Contribution
This package records a finite exact census and a scoped negative result for a four-register rational frame with exactly two independent address relations. It evaluates the proposed first-seven-prefix construction class; it does not establish an obstruction for arbitrary words or supply a new kappa.
The exact census enumerates all 120 rank-one idempotents in M4(F2), 105 off-diagonal patterns (including zero), and 196 realizable two-dimensional pattern planes. Of the planes, 176 admit a common topological order and 20 have cyclic union. The candidate screen checks all 20 exceptional planes under all six relation-signature bijections and finds no diagonal singleton endpoint. The exact 16-call positive control passes; a one-call deletion and an address corruption are rejected.
Pins
0b1cccf86e2f5132ed286a26ef0a681a495740ea10b9e340fd8e354a37a0c4813c90e5de1dd42044b5f6aca4a3fc5a42b685fa32690de78db74c1ba59a4092b8f7b9f8f54c5d53d3650d9e8d78448cdddda18321research/two-relation-pattern-census-seed52/SOURCES.jsonresearch/bit-one-relation-seed04/Reproduction
Expected markers include
PASS_TWO_RELATION_PATTERN_CENSUS,PASS_PACKAGE_METADATA_AND_PINS, andPASS_RESEARCH_BASELINE_BINDINGS_ONLY. The baseline confirms that the complex supplier still binds the selected conditional kappa; this scoped screen gives no kappa improvement.The searched family, full stopping rule, proof-relevant certificate, negative controls, exact source pins, overlap evidence, attribution, and open follow-up questions are included in the package README and task metadata. Prepared with substantial OpenAI Codex assistance; no independent human review is claimed.