High-Stakes Generality · umbrella · 2026-05-30

Transfer tests, walls, and review gates.

Sundog tests whether signatures that work in one domain survive substrate changes, and publishes the walls as receipts. Six lanes sit on this matrix: Navier-Stokes C1, Riemann, Yang-Mills, P-vs-NP, ARC-AGI, and Three-Body 15C. Navier-Stokes C1 is the strongest current witness — review-gated, with a two-regime Reading-2 witness. The rest are bounded-null, review-gated, or cost-held — also receipts, also not failures. None are theorem claims. This page is the asymmetric snapshot.

A smith who trusted her gauge at her own bench crated it out to six strange shops — the foundry and the shipwright among them — with the same note nailed to the lid: measure what you already know, and write me where it lies. Some wrote back numbers. One wrote "it bends near the furnace," and that was the letter she hung above the bench — a gauge is worth exactly the list of rooms it is known to fail in. A parable, not a result. The six lanes and their receipts are below.

High-stakes generality status matrix: six project cards (Riemann, Navier-Stokes C1, P-vs-NP, Yang-Mills, ARC-AGI, Three-Body 15C) with review-gated, cost-hold, execution-hold, and running badges plus per-project stat readouts. — Status matrix — six lanes, current verdicts and walls. Cards click through to per-lane ledgers below.

The Lanes

Six transfer tests; each one names its wall.

Each lane is a pre-registered audit chain: constants and outcome categories registered in code and spec before any row is interpreted, so the failure modes are visible before the result is. Bounded-null results are named outcomes, not silent failures.

Review gated

Navier-Stokes C1

Reading-2 witness · two Grashof regimes

Postulate-1 Reading-2 regime-2 witness now replicates at G=200 and G=300 under a portable objective. Dimension-robust (d=32 / d=18), 942,834 G=300 witness pairs, composed at matching ε_K. C1 remains UNPROMOTED until an external PDE reviewer pass closes promotion criterion (c).

Open the deep dive → Ledger Lit-pass memo

Review gated

Riemann

3 lanes, 3 bounded nulls

Three lanes returned bounded nulls by distinct substrate causes; no structural-zero edge was found in the RH-adjacent zero data. The external-review packet is filed; public surface blocked on review return. 0 claimed RH progress.

Ledger Lit-pass memo External-review packet

Review gated

Yang-Mills

powered test, informative null

An audit-first escalation: four pre-registered Wilson-target probes on SU(2) 3D 12³ landed YM-P2-NEG-A, but a follow-up audit showed those targets were not powered. So the lane escalated to a powered (split-half ICC 0.965), disjoint finite-temperature Polyakov (confinement order-parameter) test — and the compact small-loop signature still showed no rank-locality beyond controls. An informative bounded null. No mass-gap, confinement, continuum, or Clay-problem claim.

Ledger Lit-pass memo Review packet

Cheap check, spoofed, repaired

P-vs-NP

cheap to check → spoofed by drift → consensus repair

The verifier reached bounded positive at v6 on a frozen op-count gate — checking costs fewer ops (527,297) than finding (555,876), ratio 0.948587. Transplanted to the mesa bridge it was spoofed by one drifting seed block at 6.25% of budget (v0, falsified); a 3-of-4 consensus repair closed the spoof but quarantined on a block-unstable disclosure flag (v1). The bounded alignment-verification version of P-vs-NP, not the Millennium Clay problem. Its soundness and lossiness are machine-checked in Lean (axiom-clean, referee-free); the decoding hardness stays imported. The same public Lean method repo now also carries real-analysis shadow decay and its general characteristic-function law, halo minimum-deviation geometry, an Aharonov-Bohm gauge-invariance core, a machine-checked 3SAT-to-decoding Karp reduction that anchors the imported decoding-hardness to 3SAT (reduction correctness only), a finite audit game pairing a cheap full-access audit with ∀-verifier blindness, the tropical / piecewise-linear core compiling cancellation-free circuits exactly to ReLU nets at linear gate count ≤ 4N via a sharing-aware DAG, the N-1 monotone depth-as-computation theorem (region-polynomial in the cancellation-free fragment), and a seven-instance find/check ledger (max-flow/min-cut, König, 2-SAT, Pratt primality, shortest-path, syndrome, ReLU gate-count) under one Certifies interface — cheaper to check than to find, with finding imported in every instance. Full lane at /algorithmic-approximation.

Spectacle Ledger Machine-checked core Receipts

Execution hold

ARC-AGI

absolute fibers sparse; relative test held

Seven exact-grid floors stand; expanded absolute fibers stayed sparse and surfaced oracle leakage on the 108-task register. The relative-locality lane stays held until tooling admission, leak checks, result path, smoke fingerprint, and freeze marker are all filed together.

Ledger Pre-registration folder

Running

Three-Body Phase 15C

6 / 12 shards logged

The Phase 15C counterfactual horizon audit is implemented and smoke-passed; 6 of 12 lock shards are logged, but no candidate-split branch read exists yet. The Three-Body roadmap is the sibling source for Navier-Stokes Reading-2 — the same audit-chain discipline applied to a different substrate.

Ledger Workbench

Transfer-lanes diagram: per-project question, current read, and blocker columns. These are not victory cards; they are the current transfer questions and the gates that keep each result honest. — Transfer lanes — what changed, what happened, what waits. The blocker column is where the discipline lives.

Where This Sits

From a first equation to transfer tests.

The site now has two public math layers. h-of-x is the narrow geometry handle: write the inverse, name its eligibility boundary, and keep the failure case visible. High-Stakes Generality is the same discipline after the substrates get harder: ask whether a signature survives a regime change, publish the bounded nulls, and keep review gates attached to the claim.

Shadow Faraday is the calibration receipt for this style before the substrates get hard: a known classical-EM identity, a local gauge-invariant shadow, a Branch A zero-out, and named quarantines at the boundary. It is not counted as a high-stakes lane because the mathematics is already understood; it is the clean pattern that makes the Navier-Stokes, Yang-Mills, Riemann, and P-vs-NP pages legible.

Use Geometry for the visual/math shelf, Alignment for same-information comparator language, Legend for vocabulary, and Repo Map when you want the whole public route in one place.

The per-lane verdicts above are catalogued by failure mode — marginal, numerical wall, bounded null, vacuous, cost-bounded, convergence-to-null, conditional, deflationary — in the public failure-mode vocabulary, the deployed companion that names why each lane sits where it does and routes readers back to the public ledgers on this page.

Claim Boundary

Not a Clay-problem claim.

None of these six lanes claims progress on a Clay Millennium problem. Navier-Stokes C1 is a Kolmogorov-flow cell witness; Yang-Mills is a finite-lattice lane whose strongest test is an informative bounded null with no mass-gap claim; Riemann returned bounded nulls.

Not a theorem-claims page.

Receipts, witnesses, and bounded nulls are the artifact class. Reopening any lane to a theorem-facing claim requires the corresponding ledger's promotion criteria to all close.

Not symmetric weight.

Navier-Stokes C1 has its own page (/navierstokes) because it is the strongest current lane. The other five are honest cards on this matrix. Symmetric layout when the underlying results are asymmetric would be a credibility loss.

Not a launch surface for the underlying ledgers.

Each per-lane ledger remains the source of truth. This umbrella is the discipline-frame; the load-bearing claims live in docs/SUNDOG_V_*.md.

Inspection Trail

Legend — failure-mode taxonomy Shadow Faraday Zero-Out — calibration receipt Navier-Stokes C1 deep dive Sundog v Navier-Stokes Ledger Navier-Stokes Lit-Pass Memo Sundog v Riemann Ledger Riemann Lit-Pass Memo Sundog vs Yang-Mills Ledger Yang-Mills Lit-Pass Memo Sundog vs P-vs-NP Ledger P-vs-NP Lit-Pass Memo Sundog vs ARC-AGI Ledger Sundog Three-Body Roadmap Operator TODO — Current Holds At A Glance Claims and Scope h-of-x first-equation page Geometry shelf Alignment comparator primer Repo Map Legend — generality vocabulary