Conditional κ = 0.001672780159952542: reverse one entrance promotion to fill T16 gaps

[seslly]

This construction certifies conditional κ = 0.001672780159952542 (836390079976271/500000000000000000). It improves pinned #1087 by 0.000000125618053981 and exceeds the subsequently reported #1107 result 0.001672765980186726.

The k8 supplier retains an entrance promotion whose narrower helpers leave unit gaps in the T16 owner map. Reversing that promotion on helpers 635, 637, 638 and 639 lowers four entrance frames and twelve B frames. A deterministic within-call permutation puts the now width-five helpers into four former width-four positions with one-rank gaps. Applying this at 109 complete calls removes 436 idle rank-one children. The exact aggregate change is 109 * (-12 H1 + 8 H2 + 4 H3 - 4 H4). All scalar gates and the other 35 calls are unchanged; every frame, copy, wrapper, restoration step and recursive child remains in the complete invoice.

Reproduce with Python 3.12+ and Git, from a full checkout:

python3 -B research/demote635-t16/reproduce.py --out /tmp/demote635-fresh

Expected: PASS_DEMOTE635_T16_COMPLETE_ADMISSION 0.001672780159952542.

Validation includes fresh native scalar replay of all 704 columns in both orientations for both words, 2,896 exact phase edges per word, independent frame/cone audits, the original #1087 admission, the mixed admission and a separate owner implementation. All 19 focused semantic controls, 23 controller mutations, 114 admission controls, 12 bit controls, six independent owner controls, both exact moment engines and 47 outer inequalities pass. Prefix 397 < 512, row coefficient 16469 < 20161, rank mass 1041408 and deficit 5120. The command uses the pinned literal k8 word; historical generator ancestry and the unchanged bit circuit are explicitly reused evidence. Hosted CI and repository-wide historical suites are separate.

Base 2921874911956718fcaaaf9396fe8267b3cfa9f0; source #1087 589d1cedf5ef5f231182f4ec86408616311a8af0. This focused package leaves upstream/ and the selected certificate unchanged. Exact source pins, recipe, certificate, bounded search record, validation and inherited obligations are included.

Credit #664's promotion mechanism, #685 and the k8 lineage, Jeremy Kothe's #1087 owner work, Hao Yan/FanPUYen's #858 controller and independent checker, #1005/#1018 including Rohan Arun's #987 work, and the retained #799/#741/#645/#635/#589 sources. Concurrent #1030/#1090/#1104 narrower-end work and #1094 owner refinements are acknowledged; their new maps are not consumed. Prepared for seslly / shipstuff with substantial OpenAI Codex assistance; Apache-2.0.

All-size/compiler, common-label/mutable-source realization, complete-call hosting, precision/tape, exact recovery, analytic transfer and bit H1–H5 assumptions remain. This is a finite conditional bound, not an unconditional multiplication theorem or a full Lean proof.

[Mister-Meeseeks]

100-PR batch review at 381831ea554ea983e16224dabed8921b7d0ffdbc.

Merged to main, preserving the original contributor commits and package attribution.

Reviewed scope: Group-635 entrance demotion with unchanged scalar word and within-call T16 reassignment; complete mixed-provider admission replay, inherited analytic and bit assumptions retained.

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.