Passes 3064–3072: noisy D4 belief machine, edit indexing, and adaptive sensor basis

[wilcompute]

Scope

Executes the five requested fronts and two outside-box fronts, after reconciling the parallel frontier through Pass 3063. The packet was re-homed from a conflicted provisional namespace to Passes 3064–3072 before publication.

Five executed fronts

1. Noisy adaptive D4 inference: exact two-step Bayes policies on all 1,436 post-base collision classes for three explicit synthetic D4-plus-erasure likelihood models. The base collision class is assumed exact; a fully noisy base filter remains open.
2. 27 versus 28 fixed schedule: imports and independently verifies the exact 28-row full-D4 construction, then emits a proof-producing 164,220-clause central-r² SAT instance. The result remains fail-closed until an UNSAT proof is independently checked or a SAT model passes all 48,826 D4 hypotheses.
3. Edit-aware synchronization: proves the isolated cyclic block has minimum phase-edit distance two, then uses the ordering of the same three pilots as a six-symbol index alphabet. Pilot-order score is 1/2; combined omission/order score is 3/5; extra optical modes remain zero.
4. Adaptive hardware: synthesizable signed three-score likelihood accumulator and posterior-gap stopping core, with decisive, erasure-escalation and horizon testbench cases.
5. Predictive causal states: exact Moore-machine quotient 1436 -> 457, 470 total recursive states, 11-to-9 fixed-bit reduction, and 1.078290855 conditional bits saved under the frozen sparse prior.

Two outside-box fronts

• Nonabelian Fourier belief engine: conjugation-invariant D4 convolution propagates through four character gains and one scalar two-dimensional block. The nonlinear Bayes update remains explicit.
• Measurement-basis portfolio: the action becomes (triangle, sensor alphabet). In the moderate synthetic channel, the conjugacy-class detector retains 99.6683% of the full-D4 detector's one-probe risk reduction.

Paper overhaul

Adds an idempotent four-front-door integrator for w33_paper.tex, photonic_holonet.tex, holonet_machine_blueprint.tex, and docs/index.html. The overhaul corrects three stale readings: the fixed upper bound is 28 rather than 29; substitution strength is not insertion/deletion correction; logical predictive entropy is not measured heat. It cross-cites, without absorbing, the parallel Passes 3020–3024 efficiency/rank-three results and the 3060–3063 substrate corrections.

Merge gates

Do not merge until the focused workflow has regenerated all certificates, passed the RTL simulation and synthesis checks, materialized all four front doors idempotently, and compiled all three PDFs. The size-27 SAT proof is a separate long-running gate and is not falsely treated as complete. No laboratory optical, edit-rate, FPGA timing, reset-energy, or physical-heat claim is made from source completion.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

[wilcompute]

Execution note and evidence boundary:

• The provisional 3025–3033 numbering was discarded after master independently reserved 3025–3031. The published packet is canonically Passes 3064–3072.
• Exact/source-complete fronts: noisy collision-conditioned Bayes models, cyclic edit obstruction and pilot-order construction, 1436→457 future-action quotient, five-channel D4 Fourier predictor, measurement-alphabet frontier, verified 28-row D4 upper bound, adaptive RTL source/testbench, and four-front-door integrator.
• Focused evidence run 30868936926 is currently pending runner allocation. It must reconcile current master, regenerate certificates, pass focused pytest, simulate/synthesize the posterior core, prove integration idempotence, and compile all three PDFs.
• The global size-27 fixed-schedule decision is a separate proof-producing SAT workflow. Until an UNSAT proof is independently checked or a SAT model passes all 48,826 full-D4 hypotheses, the exact fixed bound remains [23,28].
• Do not merge this PR or claim FPGA, PDF, laboratory likelihood, edit-rate, physical heat, or optimum-28 evidence before those gates are observed.