Propose conditional κ ≈ 0.001729737530919310 via T148 carrier skips (full replay pending)

[dataisfire]

Proposed result

Compose the published PR #3028 T148 allocation with the PR #3042 six carrier/pivot skips. The proposed conditional point is

κ = 0.001729737530919309734204 - 10^-46 - 10^-70,

an exact gain of 1364531393747392/10^24 (approximately 1.364531393747392e-9) over the full point reported in #3042. This comparison is against that pinned predecessor, not a claim to the current record or selected frontier.

Verification status — full release pending

Local python -B research/t148-carrier-skips/verify.py --math-only passed with PASS_FINITE_T148_CARRIER_COMPOSITION_NOT_A_RELEASE. It checks all 543,456 literal owner roles, exact arithmetic, authenticated words, restore equality, frame tariffs, cleanup and rejecting controls. Full registered verification has not passed: the completed PR3030 baseline and complete new native/admission/outer release remain pending. Local full replay exhausted resources; GitHub rejected Actions dispatch despite repository permissions reporting enabled. Published now at the contributor's request as a partial contribution, with no completed-release receipt or certified improvement claimed.

The replay retains the completed-baseline gate and original scientific predicates. Both changed native providers must run freshly; sealing and finalization require both actual completed releases. The dedicated workflow and ci_full.py prepare that replay. No selected-result change is included.

Scope and attribution

The new work is research/t148-carrier-skips and its dedicated verification workflow. The branch also carries exact predecessor source packages required by its pinned dependency closure; these are inherited contributions. Credit dcharb78/#3042 for the skips, Nick Petalas/#3018/#3022/#3028 for T148 packing and release machinery, Jeremy Kothe/#3015 for R1 tooling, Marius-Paul Dumitrean/#2979 for the joint words, and the retained native/owner/outer lineage named in the package. Substantial OpenAI Codex assistance is disclosed.

The candidate retains a positive fixed-inventory cap gap. All inherited compiler, dirty-state/native realization, physical, precision/tape, caller and analytic-transfer hypotheses remain. Raw machine logs are excluded; public predecessor material is retained with provenance.

Next step: complete the mandatory baseline and candidate replay, then seal authentic receipts and update this same PR with the actual outcome.