Passes 2917-2923: seven-front Holonet breakthrough and blueprint correction

[wilcompute]

Scope

Execute the five live engineering/research fronts plus two deliberately high-risk structural falsifiers, while continuously improving—not merely appending to—the live holonet_machine_blueprint.tex.

The branch workflow first merges the newest origin/master, so every theorem, hardware, and PDF gate runs against the latest parallel work rather than the branch's creation snapshot.

Pass 2917 — seven-bit rank-coded frame engine

• Synthesizable radix-three rank engine for all 81 frames.
• Exhaustive Python and SystemVerilog closure over all 324 state/opcode transitions.
• Same external harness as the measured 43-LC arithmetic engine.
• Boundary: the seven-bit storage theorem is exact; area and timing remain pending until observed Yosys/nextpnr evidence.

Pass 2918 — complete two-copy M36 first-order census

• Exhausts 5,355 stabilizer codes, four syndromes, and the four Clifford classes.
• Deep class: 3,393 closed branches; exactly 48 improve linearly.
• Exactly 12 attain the global minimum slope 2/3; the other 36 have exact slope 2 - 2*sqrt(3)/3.
• Other three classes have minimum slope 1.
• Accepted-output mixtures are success-weighted convex combinations, so randomization or syndrome-conditioned mixing cannot beat 2/3 or cancel the linear term.
• This repairs an over-strong blueprint inference: no-quadratic alone did not prove the 2/3 optimum.

Pass 2919 — phase-sensitive separator for the two middle classes

• Complex conjugation swaps the two 12-ray classes and fixes shallow/deep classes setwise.
• Complete stabilizer-overlap probability spectra remain identical.
• Odd-Y Pauli expectations reverse sign; even-Y expectations agree.
• The operational distinction is antiunitary chirality relative to a fixed Pauli frame, not another fidelity statistic.

Pass 2920 — exact noisy adaptive diagnosis and thermodynamic repair

• Rebuilds the depth-4 adaptive support observer with uniform mean 94/27 operations.
• Exact repeated-count MAP evaluation under four asymmetric independent detector channels; no Monte Carlo.
• Three samples per decision reduce modelled error from 12.58% to 0.765% for symmetric 2% flips and from 27.94% to 2.48% in the loss-dominated stress model, at triple raw readout cost.
• Corrects the blueprint's Landauer wording:
  • 8/3 bits is H(frame|support), the discarded phase entropy;
  • H(support)=3.673183 bits is the support-record entropy;
  • log2(81)=6.339850 bits is a compressed exact identification record;
  • naive adaptive four-bit snapshot reset costs 484/27 bits on average.

Pass 2921 — live blueprint/PDF release

The fail-closed migrator:

• replaces the global two-copy slope argument with the actual first-order census;
• inserts the antiunitary chirality theorem;
• repairs the representation-dependent Landauer explanation and the compute/readout scope;
• distinguishes internal rank storage from support-observer codes;
• refines the 188-element worst-shell section;
• adds the line-stabilizer falsifier;
• preserves later parallel pass frontiers and refuses to run if any canonical anchor drifts.

The dedicated workflow double-compiles the newest-master blueprint with Tectonic, rejects TeX errors and overfull boxes, requires byte-identical PDFs, and freezes the source/PDF/results on the PR head.

Pass 2922 — outside-box line-stabilizer falsifier

The tempting conjecture fails decisively:

• the 1,296-element setwise stabilizer of the distinguished classical line is transitive on all 36 magic points;
• the 54-element pointwise stabilizer has four 9-point coordinate-family orbits;
• neither yields the Clifford 4+8+12+12 split.

Therefore the Clifford classes require extra complex/Pauli phase structure; finite geometry plus a marked line is insufficient.

Pass 2923 — outside-box algebraic census of the 188 hardest compiler elements

Full directed BFS over all 4,199,040 affine transformations:

• diameter 19, shell size 188;
• 25 exact algebraic profiles;
• affine orders 4:110, 5:6, 6:2, 8:22, 9:2, 10:6, 12:32, 18:8;
• fixed frames 0:12, 1:174, 3:2;
• inverse depths 16:18, 17:46, 18:48, 19:76.

Only 76 hardest elements have hardest inverses, so the directed worst shell is strongly asymmetric. The 25 profiles are not promoted to conjugacy classes without a separate conjugacy calculation.

Evidence already frozen

• Compact all-green seven-front summary.
• Seven normal readable verifier programs plus dispatcher.
• Synthesizable rank engine and exhaustive RTL testbench.
• Focused regressions.
• Fail-closed blueprint integrator.

Merge gate

Do not merge until:

1. the workflow merges the newest master cleanly;
2. all seven exact verifiers rebuild;
3. observed RTL simulation and rank-vs-baseline P&R are frozen;
4. the newest-master blueprint compiles twice byte-identically with no overflow/error;
5. the bot pushes the PDF/certificates/merged-master reconciliation to the PR head; and
6. a follow-up drift-free run is green.

[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]

PR #224 is the strict superset continuation of this packet. Keep #223 open and unmerged until #224's dedicated Passes 2953-2959 workflow has: (1) merged the newest master into its head, (2) rebuilt the exact certificates, (3) frozen the observed rank-engine P&R verdict and current-master Holonet PDF, and (4) completed a drift-free rerun. After those gates, close #223 as superseded rather than merging both blueprint migrations.