Skip to content
Adjacency Theory markAdjacency Theory

Checkpoint · August 26, 2026

Research Checkpoint — End of Session

A durable record of where the program stands, written so the next session can resume exactly here without relying on anyone's memory. Everything below is a property of our own equations. Nothing below is a measurement.

Scientific integrity status

  • NO PHYSICAL MECHANISM. Nothing in this program proposes or identifies a physical process.
  • NO PHYSICAL EVIDENCE. Every number is a property of equations we wrote ourselves, integrated on small synthetic systems.
  • TOY COMPUTATIONAL PROGRAM. Results are labelled MATHEMATICAL PROPERTY, COMPUTATIONAL RESULT or MODEL HYPOTHESIS, and unmodelled constraints are marked NOT MODELED rather than PASS.
  • Graph-distance reduction is never presented as spacetime manipulation. Failed experiments remain visible and versioned.
  • CURRENT STATUS: FIXED SMOOTH EMERGENCE INSUFFICIENT. The programme is a mathematical bound and a computational framework. No physical mechanism, no physical evidence.
  • Mechanism-of-the-hour branches remain preserved and rerunnable but are no longer the front of the programme. They are all downstream of the Lipschitz bound (B52).

Core objective

Determine whether a lawful, substantial change of operational relational location is possible without destructive traversal: identity and its operational invariants preserved throughout the transition rather than only at the endpoint, interaction local with bounded propagation, finite accounted action, no hidden state debt, and robustness, scale-consistency and nontriviality all demonstrated rather than assumed.

Range: AT-0 → AT-0406 (through Phase 47, the foundational reset, canonical AT-0401 … AT-0406 / requested AT-0373 … AT-0378; Phase 46 resonant adjacency is canonical AT-0391 … AT-0400 / requested AT-0363 … AT-0372)Seeds: seed 223 · seed 247 · seed 255 · seed 263 · seed 273 · seed 283 · seed 293 · seed 303 · seed 313 · seed 331 · seed 343 · seed 353 · seed 363 · seed 373Checkpoint v2.1.0

Current mathematical formulation

  • Relational address R_X = (r_X1, …, r_XN); relational distance D(A,B) = F(R); adjacency threshold D(A,B) ≤ ε.
  • Adjacency cost C_A = Σ |R'_ij − R_ij|; adjacency gain Γ_A = |ΔD| / C_A, with zero cost handled explicitly.
  • Identity is a vector observable I with a coupling map M from address change into identity change; coupled dynamics Î = I + κ·M R.
  • Scores are pathwise: peak_t |ΔI| over the whole transition, not the final reading; memory residue is read after interaction is restored to baseline.
  • Every coupling schedule obeys a mandatory floor and must restore full baseline coupling before the run ends, so no result can be bought by decoupling.
  • Auxiliary control terms must be local, bounded, frozen after discovery, necessary by ablation, and fully charged for their action.
  • Two-layer architecture (AT-0263): a deeper relational layer Y, a projection pi: Y → X onto emergent observable geometry, operational observables O, identity read only through O ∘ pi, and transitions restricted to bounded local Y steps.
  • Projection leverage L = D_X(pi(yA), pi(yB)) / D_Y(yA, yB), always reported after operator-norm (linear) or peak local Jacobian (nonlinear) normalisation and against matched random-map controls.
  • Continuity no-go (AT-0273): for a fixed smooth pi and a continuous path gamma in Y, pi . gamma is continuous in X. A skip therefore requires Jacobian stretch, singular or discontinuous behaviour, or a dynamical pi / g_X. If pi is globally 1-Lipschitz then D_X <= D_Y and large leverage is impossible. Model theorem, not a physical law.
  • Foundational bound (AT-0401): if X is obtained from S by a FIXED, single-valued, globally L-Lipschitz map F, then d_X(F(a),F(b)) <= L d_S(a,b). Normalising L ~ 1, arbitrarily large 'far in X while close in S' leverage is impossible. Model theorem plus seeded numerical check, not physics.
  • Escape classes (AT-0402), exhaustive because each negates one hypothesis of that theorem: E1 non-Lipschitz / singular or critical regions; E2 dynamical map or emergent metric; E3 lawful path metric differs from the static state metric; E4 quotient, branching or topology change. No fifth branch is admitted.
  • Dynamic layer: pi_lambda and g_X(lambda) with lambda a tracked degree of freedom. Every edit is rate-capped, propagates one link per step from the traveller's own site, is charged at a declared price, and must be restored to baseline before the run is scored.

Latest package — AT-0401 → AT-0406 (req. AT-0373 → AT-0378)

AT-0401 (req. AT-0373)

Lipschitz No-Go

For a fixed globally L-Lipschitz map F: S -> X, d_X <= L d_S by definition of the Lipschitz constant. A seeded sweep at seed 373 over random normalised linear maps in dimensions 3, 5, 8 and 12 tested 9,600 state pairs; the largest observed ratio was 0.99982 and there were no violations. MATHEMATICAL PROPERTY plus toy numerical check (B52).

AT-0402 (req. AT-0374)

Escape-Condition Audit

Exactly four escape classes, one per hypothesis of the theorem: non-Lipschitz or critical regions, dynamical map or metric, lawful path metric different from the static state metric, and quotient / branching / topology change. No speculative fifth branch admitted (B53).

AT-0403 (req. AT-0375)

Cost of Escape

Each escape class carries named debts already in this ledger: singular Jacobian and identity amplification, map-edit action and front precedence, pathwise causality and route specificity, topology bookkeeping and gauge discipline. Three of four are priced and all cost more than ordinary traversal; topology change is NOT MODELED and therefore unavailable (B53).

AT-0404 (req. AT-0376)

Same-Equations Dual-Regime Test

One minimal family f_a(x) = x + (a/k) sin(kx), one parameter, no hand-added channel. Ordinary finite-speed emergent behaviour appears at every setting; the apparent leverage at high sharpness is entirely growth of the global Lipschitz constant and the L-normalised stretch never exceeds 1. No setting gives both regimes (B54).

AT-0405 (req. AT-0377)

Blind Emergence Search

Generate, freeze, then search, in that order in code, with the shortcut objective never consulted during generation. Across 24 frozen models and 5,760 endpoint probes the best straight-line leverage was 0.962 and the best lawful leverage 0.954. Nothing above the bound occurs naturally. NOT EXPECTED BY CONSTRUCTION, and still negative.

AT-0406 (req. AT-0378)

Foundational Gate

FIXED SMOOTH EMERGENCE INSUFFICIENT. Five of ten criteria pass; every failing criterion is substantive. Next work must identify which escape condition is mathematically necessary and whether it can arise from the SAME underlying equations without hand insertion. No new mechanism family may be opened until that question is posed as a theorem.

AT-0391 (req. AT-0363)

Adjacency-as-Primitive Baseline

Bond A_ij(t) treated as the primitive dynamical variable with distance derived only afterwards. The seed-363 null reproduces the recorded statistics and is kept as a control, never a finding: the compatibility that drives the bond was inserted into its own equation (B49).

AT-0400 (req. AT-0372)

Resonant Adjacency Gate

Gate not cleared. Bonds with no similarity hand-off grow everywhere at once, tuned bonds need continual retuning and do not transfer, addressing costs about ten bits against one bit of payload, and the full activation ledger costs more than ordinary traversal (B50, B51).

AT-0381 (req. AT-0353)

Local Reservoir Ledger

Every edge of a 60-edge route holds a stored resource; each transit draws 0.8 per edge and bounded local recharge puts a little back between trips. Eight trips complete on a loading of 360 units, and the reservoir balance closes to numerical zero — nothing appears without a source.

AT-0382 (req. AT-0354)

Full Amortized Cost

Marginal external input after preload is 4.40 per trip. Fully amortized — preload, recharge, triggers, maintenance, less what is left in the route — it is 49.40 per trip, against 48.00 for ordinary traversal. An eleven-fold understatement (B46).

AT-0383 (req. AT-0355)

Route-Length Scaling

Preload grows at 6.00 units per edge with R² above 0.999 and per-trip cost tracks it, while the trigger stays flat in length. Infrastructure energy is linear in route length; there is no length at which preparing the route becomes cheap (B47).

AT-0384 (req. AT-0356)

Utilization Break-Even

A one-off dynamic corridor never amortizes; the preconditioned route beats it quickly but edge capacity caps a loading at eight completed trips, so it does not cross ordinary traversal in the tested range. Any crossover exists only inside these toy assumptions.

AT-0385 (req. AT-0357)

Reservoir Propagation

Delivering the loading locally from a depot costs transport overhead that grows with distance, plus charged setup time. Instantaneous placement everywhere is priced and then rejected as a forbidden control — it is the assumption that preparation is free.

AT-0386 (req. AT-0358)

Destination Independence

A corridor prepared between two sites is reused by only a small fraction of a randomly requested journey, and most journeys get nothing from it. Classified ROUTE-SPECIFIC INFRASTRUCTURE, not universal adjacency (B48).

AT-0387 (req. AT-0359)

Network Geometry

Prepared chords on a ring cut the mean pairwise cost and produce a real effective-adjacency gain, entirely by graph routing over links somebody paid for. Energy per unit of gain rises as chords are added.

AT-0388 (req. AT-0360)

Hub Hypothesis

Preparing every pair is O(N²) and gives the shortest journeys; hub-and-spoke costs far less and stretches every journey through one node; the hierarchy sits between. The standard cost-versus-stretch tradeoff of any transport network, with no physical content.

AT-0389 (req. AT-0361)

Universal-Medium Challenge

A medium homogeneous enough to serve arbitrary endpoints costs with the area it covers rather than the journey it saves, many times a single corridor. All negative controls fail as required: unprepared carries nothing, and the same energy spread below the activation draw is stranded.

AT-0390 (req. AT-0362)

Infrastructure Gate

Recorded status: GATE NOT CLEARED, classified ROUTE-SPECIFIC INFRASTRUCTURE. Maximum award available remains STORED-ENERGY INFRASTRUCTURE TOY / PHYSICAL EVIDENCE NONE.

AT-0371 (req. AT-0343)

Phase-Switch Amplification Audit

A tiny trigger in a single-field bistable medium produces an integrated response several hundred times its own energy, but the field settles into a persistent alternate state and no candidate restores near baseline. Classified PHASE-SWITCH AMPLIFICATION WITH STATE DEBT (B43).

AT-0372 (req. AT-0344)

Recovery-Field Baseline

With an explicit recovery field the medium restores to between about 1e-8 and 1e-4, and the excitation never leaves the trigger neighbourhood — a front of only a few cells. TRANSIENT LOCAL AMPLIFICATION / NO LONG-RANGE PROPAGATION (B44).

AT-0373 (req. AT-0345)

Propagation-vs-Restoration Tradeoff

Across the excitability-recovery plane the regions are disjoint: cells either restore and stay local, or propagate and leave the medium switched. Zero cells did both (B44).

AT-0374 (req. AT-0346)

Stored-Medium Energy Ledger

Every cell carries an explicit free-energy reservoir; positive reaction work is only permitted where the reservoir can pay. Reach scales with stored energy and the debit matches the work to numerical tolerance (B45).

AT-0375 (req. AT-0347)

Trigger-vs-Reservoir Scaling

Trigger cost is flat in route length; preloaded infrastructure energy and per-transit depletion are linear in it. A finite trigger covering a long route means a long route was charged in advance (B45).

AT-0376 (req. AT-0348)

Preconditioned Medium Test

Prepositioned infrastructure is admitted as its own mechanism class. Amortised cost per trip falls with reuse but the setup energy is never dropped, and in the tested range it does not undercut ordinary traversal.

AT-0377 (req. AT-0349)

Self-Recharging Medium Challenge

Closing the cycle requires explicitly accounted external recharge at bounded local rate, plus waiting time. Nothing in this family recharges itself.

AT-0378 (req. AT-0350)

No-Free-Amplification Control

Under the declared energy functional, response work never exceeds trigger plus reservoir debit plus external input in any sampled configuration. Any positive residual is reported as a model bug, not a discovery.

AT-0379 (req. AT-0351)

Infrastructure-vs-Vehicle Split

Three accounting models — stored along the route, carried by the traveller, supplied from origin — are reported separately and never summed. Switching between them silently is how a shortcut gets manufactured.

AT-0380 (req. AT-0352)

Stored-Medium Gate

Recorded status: GATE NOT CLEARED. Maximum award available remains STORED-MEDIUM CORRIDOR IN TOY MODEL / PHYSICAL EVIDENCE NONE.

AT-0361 (req. AT-0343)

Multi-Observer Invariant Audit

Independent charts of the same run agree on event order and on normalised displacement to machine precision (~1e-14). Hygiene met, no evidential weight (B42).

AT-0362 (req. AT-0344)

Local-Scale Power Ceiling

Under bounded amplitude, |dλ/dt| < 3, a gradient bound and enforced restoration, best gated leverage is about 1.5 at a large accounted action; most gated candidates sit between 1.3 and 1.5 (B40).

AT-0363 (req. AT-0345)

Nonlinear Scale Interaction Search

A bounded local self-interaction is necessary for the gain — the linear ablation is measurably weaker — and the Jacobian guard confirms no gain was manufactured near singularity.

AT-0364 (req. AT-0346)

Critical Scale Response

No phase-transition-like response found: the reorganised region grows in proportion to the trigger, with denominator guards active near zero baseline.

AT-0365 (req. AT-0347)

Self-Amplifying Scale Pulse

A finite local trigger spreads by diffusion and decays, retaining only a few percent of its amplitude. Bounded self-propagation NOT demonstrated; never called a soliton.

AT-0366 (req. AT-0348)

Scale-Pulse Backreaction

With the traveller pushing back on the field the leverage degrades slightly. The free-ride figure is retained as a control and never reported as the result.

AT-0367 (req. AT-0349)

Action-vs-Leverage Law

Action grows superlinearly with leverage across the gated set. Recorded as ENGINEERINGLY UNFAVOURABLE IN TOY MODEL; no cheap regime exists in the sampled family (B41).

AT-0368 (req. AT-0350)

Observer-Causal Consistency

No ordering contradictions. Chart-specific apparent superluminality is a bookkeeping artefact and is rejected as a claim.

AT-0369 (req. AT-0351)

Emergent-Limit Recovery v2

At baseline λ the branch reduces to the ordinary finite-speed emergent behaviour, and the field restores inside the debt tolerance. Necessary condition, not an achievement.

AT-0370 (req. AT-0352)

Scale-Field Gate v2

Recorded status: GATE NOT CLEARED. Maximum award available remains LOCAL-SCALE LEVERAGE IN TOY MODEL / PHYSICAL EVIDENCE NONE.

AT-0331

Identity Through the Travelling Corridor

G10 closed by measurement. The traveller now has an internal identity vector; riding the corridor peaks at about twice the identity disturbance of walking the same route.

AT-0332

Front-Precedence Bound

Causal cutoff enforced: the traveller is never helped by a corridor that has not reached it. Any net advantage is bounded below by the front's own crossing time (B28).

AT-0333

Matched-Progress Identity Comparison

The ride is worse than the walk at every matched fraction of the route. Median ratio about 2x. Matched progress does not rescue it.

AT-0334

Smoothness / Affordability Frontier

Peak identity disturbance and accounted action are anti-correlated across corridor width. Gentleness is purchased, never given (B35).

AT-0335

Synchronisation-Free Corridor Search

Launch-offset sweep. The tolerant window is only a few model time units wide, so the arrangement still needs an unmodelled coordinating agent (B26).

AT-0336

Standing-Medium Legitimacy Audit

Setup and maintenance are charged separately from the per-trip cost. The standing corridor is classified PREPOSITIONED INFRASTRUCTURE and never wins on time for one trip (B30, B33).

AT-0337

Geometry Price Derivation Attempt

Four pricings, and the verdict moves. The cost of a geometry edit is declared by us, not derived from the model's own dynamics (B25).

AT-0338

Closed-Cycle Identity Debt

MATHEMATICAL PROPERTY of the coupling law: the identity drive is sign-blind in the gradient, so an out-and-back cycle accumulates residue instead of cancelling it (B36).

AT-0339

Carrier Nontriviality Control

A sham corridor with identical field and identical charge but no coupling produces neither the benefit nor the damage. The result is structural in the model, not a bookkeeping artefact.

AT-0340

Carried-Traveller Gate v1

3 of 13 requirements met, 2 NOT MODELED. Recorded status: GATE NOT CLEARED, NEGATIVE RESULT / DAMAGE RIDES WITH THE CORRIDOR.

AT-0255

Constrained Schedule Search

RETUNED TOY IMPROVEMENT — median peak identity disturbance falls roughly a fifth, but every schedule is re-tuned per system.

AT-0256

Frozen Schedule Transfer

One schedule discovered on a training ensemble, frozen, and carried to unseen systems, higher dimension, sparser topology, larger displacement and heavier drift.

AT-0257

Universal Schedule No-Go / Survival

Compact fixed schedules compared against fixed coupling and an adaptive gate on identical cases with no retuning.

AT-0258

Peak-Damage Floor

A positive lower bound on achievable worst-case identity disturbance persists at matched destination error and enforced coupling floor. This is the most restrictive negative result currently standing.

AT-0259

Moving-Invariant Candidate

Redefining identity as a dynamical equivalence class lowers the score only when observable completeness falls; negative controls catch the vacuous quotients.

AT-0260

Counterdiabatic-Like Toy Control

A purely mathematical local auxiliary term reduces transient excitation with its full action charged. Toy control analogy only; no connection to physical driving protocols is claimed.

AT-0261

Control Necessity and Locality Audit

Locality, boundedness, freezing, ablation necessity and a generic-feedforward comparison are all required; direct identity freezing is forbidden.

AT-0262

End-of-Day Gate

Full blocker vector rerun with an explicit resume state. Recorded status: NONE PASSED.

AT-0263

Two-Layer State Model

Y, pi, X, O, I and T formalised. Einstein-like local behaviour is kept as an emergent benchmark, not adopted as an axiom.

AT-0264

Projection Leverage Null

Generic Gaussian maps already amplify by their largest singular value. Raw leverage is labelled EXPECTED BY MAP GEOMETRY; normalised leverage cannot exceed 1 for a linear map.

AT-0265

Nonlinear Projection Leverage

Smooth nonlinear projections relabel amplification rather than creating leverage once the peak local Jacobian is charged.

AT-0266

Short-Y / Far-X Search

Lawful bounded local Y paths searched against a matched random-map control on the identical path family.

AT-0267

Projected-Path Audit

The projected trajectory is sampled throughout. Candidates that sweep the ordinary intermediate region are rejected as traversal, not leverage.

AT-0268

Emergent Locality Test

POSITIVE: a finite propagation cone emerges from purely local Y interactions with no speed inserted. Reported dimensionlessly, in sites per integration step.

AT-0269

Dual-Channel Challenge

The ordinary channel exists; no distinct lawful high-leverage transition class survives the audits. The dual-channel question is unanswered, not answered.

AT-0270

Identity Across Projection

Identity is read only through O ∘ pi with the definition fixed before the search; the unmeasured fraction of Y is reported alongside every deviation.

AT-0271

Causal-Consistency and Signalling Audit

Toy signalling protocols constructed. No contradiction arises only because no channel outran the emergent cone; that is an absence of a candidate, not causal safety.

AT-0272

Deeper-Layer Gate

Full blocker stack plus projection-specific blockers. Recorded status: NONE PASSED.

AT-0273

Fixed-Projection Continuity No-Go

MODEL THEOREM. A fixed smooth projection maps continuous paths to continuous paths, so it cannot produce skip-the-middle relocation. Only Jacobian stretch, singularity, or a dynamical map remain.

AT-0274

Lipschitz Bound Audit

Normalised random maps at seed 273 never exceed leverage 1, and the sampled maximum falls with dimension. Leverage at or below the bound is EXPECTED BY PROJECTION STRETCH.

AT-0275

Singular-Map Trap

Near-singular and discontinuous projections built deliberately as positive controls for fakery. All rejected: leverage collapses after normalisation and sensitivity explodes.

AT-0276

Dynamic Projection Hypothesis

pi promoted to pi_lambda with lambda tracked: rate cap, propagation frontier, declared price, mandatory restoration. The only route the AT-0273 theorem leaves open.

AT-0277

Dynamic Metric Hypothesis

Object motion in fixed geometry separated from geometry change around a nearly static object. The distinction is now measurable, and the cost visibly migrates between sectors.

AT-0278

Metric-Change Cost

Geometry edits are charged and closure is enforced. At the declared price the dynamic arm costs more total action than ordinary traversal.

AT-0279

Adjacency-Without-Object-Traversal Toy

The traveller's own arc collapses and peak identity disturbance falls sharply, but the geometry edit crosses the region instead. Classified COST MIGRATION, NOT NON-TRAVERSAL.

AT-0280

Geometry-Restoration Cycle

Full cycle closes: geometry returns to baseline, ordering stays consistent, residue is read only after closure.

AT-0281

Emergent-Einstein Benchmark

POSITIVE: ordinary perturbations in the restored geometry still propagate with the same finite dimensionless front speed as before the cycle. Toy benchmark, no GR derivation claimed.

AT-0282

Dynamic-Geometry Gate

Strict stack including stretch, singularity, coordinate, cost, debt, restoration, locality and nontriviality. Recorded status: NONE PASSED.

AT-0283

Geometry Compression Cost Curve

TOY COST BENCHMARK at seed 283. Compression is easy while geometry edits are free; once manipulation, equal restoration and roughness are charged the cost-versus-shrink curve turns sharply upward.

AT-0284

Matched-Compression Efficiency

Every control law bisected onto the same shrink ratio and the same restoration accuracy. The cheapest law still costs multiples of ordinary traversal of the same chain (B22).

AT-0285

Smooth-vs-Localized Deformation

Broad smooth deformation is far cheaper in action than concentrated deformation at equal compression — but it is cheap precisely because it moves everything, which is what AT-0290 then charges for.

AT-0286

Critical Geometry Response

Apparent threshold-like gain in the nonlinear control law is a DENOMINATOR ARTEFACT NEAR THRESHOLD. Guarded, the marginal return on extra action does not spike.

AT-0287

Geometry Soft-Mode Search

Hessian eigen-decomposition with full-rank and nullspace audits. The efficiency of the softest collective mode tracks the dimension count and is EXPECTED BY DIMENSION COUNT, not emergent physics.

AT-0288

Traveling Geometry Pulse

A rate-capped deformation front shortens each link just ahead of the traveller while obeying the emergent finite speed. Local crossing is short; the control front still sweeps the region (B21). No superluminal behaviour claimed.

AT-0289

Geometry Restore Closure

After crossing, every control variable and every operational distance away from the destination must return within tolerance. Residue is recorded as debt rather than discarded.

AT-0290

Background-Disturbance Audit

New metric Collateral Geometry Distortion. Unconstrained compression moves a large fraction of unrelated pairwise distances at peak (B23).

AT-0291

Localized Corridor Challenge

Key score: useful compression divided by action plus collateral distortion plus debt. Bounded corridors suppress the background disturbance but pay a sharp premium, and narrow corridors are infeasible under the deformation cap (B24).

AT-0292

Dynamic-Geometry Gate v1

Thirteen requirements spanning genuine distance reduction, bounded local dynamics, matched efficiency, collateral distortion, pathwise identity, short local crossing, exact restoration, debt, robustness, scale, nontriviality and emergent locality before and after. Recorded status: NONE PASSED. Maximum available award ROBUST IN TOY MODEL / REQUIRES PHYSICAL THEORY; physical evidence NONE.

AT-0293

Localized vs Global Geometry Control

COMPUTATIONAL RESULT / MODEL-DEPENDENT at seed 293. At matched compression the localized corridor reproduces the recorded benchmark (Gaussian action ~36.4, collateral ~7.7%) against a global rewrite at ~101 action and ~55% collateral.

AT-0294

Corridor Width Optimization

A Pareto frontier rather than an optimum: narrow corridors keep the background still but cost more action and grow sharp shoulders; wide ones smooth the shoulders and leak into unrelated distances.

AT-0295

Boundary / Gradient Debt

The localized advantage survives a sixty-four-fold escalation of the gradient price, but only because the corridor widens to pay the boundary debt with breadth.

AT-0296

Traveling Pulse Dynamics

The corridor is grown by rate-capped nearest-neighbour updates at a bounded front speed. No remote or instantaneous geometry edits are permitted anywhere in the run.

AT-0297

Pulse-Crossing Synchronization

The traveller may cross only while the corridor is compressed. The window is finite and mistiming destroys the advantage, so the mechanism depends on coordination it cannot explain (B26).

AT-0298

Pulse Restoration

The field relaxes back within tolerance and the cycle closes, but the disturbance emitted outside the corridor during the cycle is recorded rather than forgiven.

AT-0299

Collateral Causality Audit

Local propagation reach outside the corridor changes by well under a percent and no distance ordering flips. A rejection filter passed, not a positive result.

AT-0300

Multi-Dimensional Corridor

On a square lattice the action advantage survives and grows, but the corridor becomes a tube and collateral distortion among unrelated pairs rises by more than an order of magnitude (B27).

AT-0301

Frozen Corridor Law Transfer

POSITIVE within the toy: a single scale-relative rule — corridor width about 0.40 of target separation — transfers to unseen sizes, separations, backgrounds and noise with negligible penalty and no retuning.

AT-0302

Dynamic-Geometry Gate v2

Fourteen requirements. Recorded status: 12 of 14 passed — identity coupling is NOT MODELED in this phase, and the full cycle still costs more than covering the distance outright. GATE NOT CLEARED. Physical evidence NONE.

AT-0303

Control-Front Speed Tradeoff

Sweeping the driven front speed at seed 303: faster fronts complete sooner but action rises superlinearly and collateral distortion widens. No free speed anywhere in the swept range (B29).

AT-0304

No-Problem-Transfer Audit

DECISIVE NEGATIVE. The traveller walks about 45% of the baseline distance, but the control front must reach the far target first and net completion including restoration is about 1.8x ordinary traversal. PROBLEM TRANSFERRED TO CONTROL FIELD (B28).

AT-0305

Self-Propagating Pulse

A bistable excitable medium with refractory recovery sustains a front after finite local initiation at about 0.67 links per step. Self-sustaining, still finite-speed, still dissipative.

AT-0306

Pulse Soliton-Like Stability Toy

Six of nine amplitude-width profiles survive perturbation with near shape preservation. Stability is a property of the medium, not evidence of a shortcut.

AT-0307

Pre-Existing Medium vs On-Demand Corridor

A standing medium costs about 61.5 per crossing against about 110 on demand, but only after preparing the entire chain and paying maintenance forever. Completion is still slower than ordinary traversal.

AT-0308

Bidirectional Restoration Wave

The trailing restoration wave closes the cycle within tolerance, but restoration is charged at the same rate as construction and the residue is recorded rather than forgiven.

AT-0309

Traveller-Pulse Synchronization

The crossing window is narrow: delays beyond a few steps miss the corridor entirely and the traveller never completes. The advantage is conditional on coordination that has no mechanism (B26).

AT-0310

Net Time-to-Destination

Net completion including front propagation and restoration is about 43.2 model time units against about 24.0 for ordinary local traversal — a ratio of about 1.80. The shortcut is slower than walking.

AT-0311

Infrastructure-Amortized Test

INFRASTRUCTURE-ONLY BENEFIT. Amortized over repeated crossings the medium wins on action after roughly nine crossings, but never on time, and it is a prepared channel between two prepared places (B30).

AT-0312

Travelling-Corridor Gate

Fifteen requirements. Recorded status: 10 of 15 passed. Decisive failures on net completion time and on problem transfer to the control field. GATE NOT CLEARED. Physical evidence NONE.

AT-0313

One-Shot Trigger Baseline

NEGATIVE RESULT: SIMPLE EXCITABLE MEDIUM INSUFFICIENT. A single bounded local trigger in a one-dimensional excitable reaction-diffusion geometry medium opened roughly a third of the sixty-nine-edge route and then collapsed. Zero full crossings (B31).

AT-0314

Excitability Threshold Map

Threshold, diffusion, damping, trigger width and trigger amplitude swept and classified as decay, pinned front, partial propagation, full propagation or blow-up. No full propagation anywhere in the swept box; past the boundary the medium diverges instead.

AT-0315

Soliton-Like Pulse Search

A conservative dispersive toy medium does produce localized structures that travel with no drive and restore the background after passage — and none of them carries enough compression to help a traveller. Self-propagation is necessary, not sufficient (B32). Pure mathematics, no physical claim.

AT-0316

Pulse Energy/Action Accounting

Trigger action, carried-field action and residual background distortion accounted separately and compared against the continuously driven corridor. The cheap arm is only cheap because it never arrives.

AT-0317

Traveller-Pulse Synchronization

The traveller may only move through currently compressed links. On the self-propagating arm there was nothing to walk through; the driven arm completes only under continuous remote actuation and still not faster than ordinary traversal.

AT-0318

Medium Preconditioning

One-time preparation cost is reported separately from per-trip trigger and maintenance cost, so setup debt can never be hidden inside an attractive per-trip figure (B33).

AT-0319

Reusability and Degradation

Repeated pulses through the same medium with the recovery field carried forward. Reach drift is recorded; repeatability of a failure is not counted as support.

AT-0320

Bidirectional / Reversal Test

A to B and B to A under the same frozen medium and rule give comparable reach and comparable restoration, confirming the rule is directionally symmetric rather than confirming the hypothesis.

AT-0321

Causal-Front Audit

The traveller never advances beyond the furthest point the front has reached. Any surviving net advantage traces to a medium prepared along the whole route in advance and is classified PREPOSITIONED INFRASTRUCTURE (B33).

AT-0322

Self-Propagating Corridor Gate

Thirteen requirements covering one-time triggering, finite front speed, full-route propagation, collateral, restoration, repeatability, reversal, setup debt, identity, action, net time and prepositioning. GATE NOT CLEARED, decisive failure on full-route propagation. Physical evidence NONE.

Run them at /lab/schedules.

Killed shortcuts and no-go findings

Claim that failedWhy it failedExperiments
K1Identity and relational address can simply be decoupled.The independence was built into the model by construction, not discovered from it.AT-0073 → AT-0080
K2A protected nullspace shields identity during relocation.The protection came from a dimension mismatch. With a full-rank coupling it disappears entirely.AT-0087 → AT-0094
K3Identity is preserved because the observed state is unchanged.Incomplete observables manufacture equivalence. Measuring less is not preserving more.AT-0127 → AT-0134
K4A change of relational coordinates is a change of relational location.Gauge-redundant directions divide out. Once the physical quotient is taken, the shortcut vanishes.AT-0135 → AT-0142
K5Path choice can avoid identity damage.With constant linear coupling, endpoint identity change is path-independent. Any path dependence introduced afterwards produces loop residue and memory debt.AT-0103 → AT-0118
K6Soft modes of the coupling are the cheap directions to move along.Softness is not selectivity. Soft directions were no better aligned with useful relocation than random ones.AT-0159 → AT-0166
K7Nonlinearity protects identity.Strong damping explains most of the apparent protection, and the identity definition used was too loose.AT-0183 → AT-0198
K8Suppressing coupling late in the transition preserves identity.Peak damage is essentially unchanged. Classified ENDPOINT CLEANUP / NOT PATHWISE PROTECTION.AT-0247 → AT-0254
K9A rich nonlinear cross-term family contains a survivor.Six thousand candidates produced zero strict survivors.AT-0167 → AT-0182
K10All four constraints — progress, identity, invariance and locality — can hold together.The admissible intersection was empty in the toy model. The joint wall stands.AT-0143 → AT-0158
K11A deeper layer gives large projection leverage, so a short move below is a long move above.Generic random maps already amplify distances by their largest singular value. After operator-norm or Jacobian normalisation, no linear or smooth projection exceeded L = 1. EXPECTED BY MAP GEOMETRY.AT-0264 → AT-0265
K12Far-apart projected endpoints mean the transition skipped the intervening region.Sampling the whole projected path showed it walking through the ordinary intermediate locations. Endpoint separation is not non-traversal.AT-0266 → AT-0267
K13A deeper layer can let a continuous journey skip the middle.For a fixed smooth projection the image of a continuous path is continuous. The only escapes are Jacobian stretch and singularity, and both are artefacts we now build deliberately as controls.AT-0273 → AT-0275
K14Changing the geometry instead of moving the object avoids the cost of the journey.The cost migrated rather than vanished: total accounted action across both sectors exceeded ordinary traversal, and the geometry edit itself had to propagate across every intervening link.AT-0276 → AT-0282

Current best surviving clues

Timing and coupling geometry matter more than controller complexity.

At matched relational progress, a simple two-stage scheduled gate matched or beat the adaptive feedback rule on identity disturbance.

Caveat: This downgrades adaptive control rather than promoting scheduling. It is a warning about our scoring, not a mechanism.

Constrained schedule search reduces peak identity disturbance with the coupling floor and restoration both enforced.

Roughly a sixth to a fifth of peak damage removed across sampled cases at seed 255, with the interaction never switched off.

Caveat: Schedules are re-tuned per system and remaining damage is still large. Label: RETUNED TOY IMPROVEMENT.

A local, bounded, frozen auxiliary term can suppress part of the transient excitation.

The reduction survives explicit action charging and an ablation necessity test in the toy plant.

Caveat: Purely mathematical toy control. No physical driving analogy is claimed, and it does not clear the peak-damage floor.

Finite propagation in the observable layer emerges from purely local interactions in the deeper layer.

Arrival time grows linearly with graph distance with no speed inserted anywhere in the model, giving a dimensionless emergent front (AT-0268).

Caveat: It is a benchmark our own model must reproduce, not a claim about physics, and the number has no units and is not a speed of light.

Moving the geometry rather than the traveller sharply reduces the traveller's own identity disturbance.

The traveller's arc collapses to a single short local step, and peak identity deviation drops with it, at a genuinely large operational displacement (AT-0279).

Caveat: The damage and the cost move into the geometry sector rather than disappearing, and the edit still propagates across the region. COST MIGRATION, NOT NON-TRAVERSAL.

A bounded corridor is far cheaper and far quieter than a global rewrite, and the rule that picks it transfers.

At matched compression the corridor costs about a third of the action and moves unrelated distances by about a tenth as much; a single scale-relative width rule, frozen after fitting, carries to unseen sizes and separations with negligible penalty (AT-0293, AT-0301).

Caveat: COMPUTATIONAL RESULT / MODEL-DEPENDENT. The comparison is decided by an assumed cost functional (B25), the advantage needs synchronisation (B26), and the cycle still costs more than ordinary traversal.

Blocker ledger

B52

Fixed smooth emergence forbids leverage

If emergent geometry comes from the fundamental state space by a fixed, single-valued, globally L-Lipschitz map, then d_X <= L d_S and, after normalising L to 1, arbitrarily large leverage is impossible. Every mechanism family in this project so far sits downstream of such a map, which is why they all failed in the same way (AT-0401).

OPEN
B53

Only four doors, all with bills attached

The bound can fail only by negating one of its own hypotheses: critical or non-Lipschitz regions, a dynamical map or metric, a lawful path metric that differs from the static one, or a quotient / branching / topology change. Three are priced and cost more than ordinary traversal; the fourth is unpriced and therefore unavailable. No mechanism may be proposed unless it is first identified as one of these four (AT-0402, AT-0403).

OPEN
B54

One model must produce both regimes

No arm counts unless a single underlying model, with no second channel added by hand, reproduces ordinary finite-speed emergent behaviour AND the anomalous regime. In the minimal smooth family the anomalous regime was a units choice, and blind search on frozen geometry found nothing above the bound (AT-0404, AT-0405).

OPEN
B49

Compatibility forcing is not emergence

In the seed-363 null the bond tracks a pairwise compatibility term inserted into its own equation. Ablating it collapses selective bonding and shuffling it preserves the headline rate while destroying every pairwise association. No adjacency result may be quoted before an ablate-and-shuffle audit (AT-0391, AT-0392).

OPEN
B50

Adjacency addressing bottleneck

Aiming a bond at one destination out of many needs selection bits plus a description of the remote state, about ten bits against one bit of payload, and those bits must arrive over an ordinary channel first. A bond cannot be aimed at a place you have not already heard from the slow way (AT-0395, AT-0398).

OPEN
B51

Resonance without selectivity, tuning without transfer

Bond rules with no similarity hand-off produce bonds everywhere at once; tuned bonds exist only in a narrow detuning window, need continual retuning against drift, do not transfer to another pair, and cost more than ordinary traversal once fully charged (AT-0393, AT-0394, AT-0396, AT-0397).

OPEN
B1

Trivial construction

Any benefit that is a restatement of a modelling choice is vetoed by the nontriviality gate.

OPEN
B2

Observable incompleteness

Completeness is now audited, but the audit is only as good as the channel list.

PARTIALLY ACCOUNTED
B3

Gauge redundancy

Physical quotients are taken explicitly; representational relocation is rejected.

PARTIALLY ACCOUNTED
B4

Hidden debt

Memory residue after coupling restoration remains nonzero for every mechanism tested.

OPEN
B5

Locality and bounded propagation

Controls are constrained to sparse local bands, but propagation bounds are still coarse.

PARTIALLY ACCOUNTED
B6

Action budget

Coupling changes and auxiliary control are charged; the exchange rate between them is a modelling choice.

PARTIALLY ACCOUNTED
B7

Retuning

A rule that must be re-tuned per system is not a law. Frozen transfer is the discriminating test.

OPEN
B8

Scale consistency

Coarse-grained agreement is untested for the Phase 33 mechanisms.

OPEN
B9

Transient versus endpoint

Scoring is pathwise. Endpoint identity is retired as a headline number.

CLOSED
B10

Identity vacuity

Negative controls catch quotients that improve the score by measuring less.

PARTIALLY ACCOUNTED
B11

Memory and holonomy

Closed cycles do not return the system to its starting state without residue.

OPEN
B12

Gauge-equivalent endpoints

Endpoints identical up to representation are not counted as relocation.

PARTIALLY ACCOUNTED
B13

Rank deficiency

Deficient coupling rank manufactures apparent protected subspaces.

PARTIALLY ACCOUNTED
B14

Physical versus representational address

No modelled quantity has been shown to be a physical address rather than a coordinate.

OPEN
B15

Peak-damage floor

A positive lower bound on worst-case identity disturbance persists at matched progress and enforced coupling floor.

OPEN
B16

Projection scaling

Distance amplification through a map is generic; leverage means nothing until it is normalised by the operator norm or the peak local Jacobian.

PARTIALLY ACCOUNTED
B17

Projected traversal

A projected path that sweeps the ordinary intermediate region has skipped nothing, however far apart its endpoints look.

OPEN
B18

Emergent causal consistency

Any channel faster than the emergent cone must come with a consistent deeper ordering, or it manufactures closed loops in the observable layer.

OPEN
B19

Fixed-map continuity

For a fixed smooth projection the image of a continuous path is continuous. No skip is available without making the map itself dynamical.

CLOSED
B20

Geometry-edit cost migration

Charging the geometry edit moves the cost between sectors rather than removing it; total accounted action still exceeds ordinary traversal.

OPEN
B21

Geometry-control traversal

The control change itself has to propagate across the intervening region, so the region is crossed by something even when the traveller barely moves.

OPEN
B22

Compression price

Once manipulation, equal restoration and roughness are charged, the cheapest control law achieving the target compression still costs several times ordinary traversal.

OPEN
B23

Collateral geometry distortion

The action-cheapest compressions are the broadest, and they move a large fraction of unrelated pairwise distances. A short cut that distorts the whole universe is not a short cut.

OPEN
B24

Corridor premium

Confining the deformation to a bounded relational channel removes the collateral distortion but raises the action per unit of compression sharply, and narrow corridors become infeasible under the deformation cap.

OPEN
B25

Assumed action functional

Which compression is cheaper is decided by a cost rule we wrote down rather than derived. The corridor advantage survives three reasonable pricings, but no pricing is forced by the model.

OPEN
B26

Synchronisation fragility

The corridor is open for a finite window and the traveller must arrive inside it. The advantage is conditional on coordinating an object with a geometry edit, and no mechanism for that coordination exists.

OPEN
B27

Corridor is not dimension-free

In one dimension a corridor disturbs almost nothing. On a lattice the same corridor becomes a tube and collateral distortion among unrelated pairs rises by more than an order of magnitude.

OPEN
B28

Problem transferred to the control field

The corridor does not exist until the control front arrives, and the front obeys the same locality bound as the traveller. Setting up the shortcut requires crossing the route once already.

OPEN
B29

Front-speed tradeoff

A faster control front shortens completion but raises action superlinearly and widens collateral distortion. Speed is bought, never given.

OPEN
B30

Infrastructure is not a shortcut

A pre-existing medium beats on-demand construction per crossing only after a large one-time preparation and a maintenance charge that never stops, and it still never beats traversal on time.

OPEN
B31

Simple excitable medium insufficient

A one-time local trigger in a one-dimensional excitable reaction-diffusion geometry medium failed to launch a full-corridor pulse anywhere in the swept region: fronts decayed, pinned, or the medium blew up. Scope is the medium tested, not a universal no-go (AT-0313, AT-0314).

OPEN
B32

Propagation is not usefulness

Conservative dispersive structures that travel indefinitely without drive and restore the background afterwards still carry too little compression, too briefly, to help a traveller. Self-propagation and usefulness are separate requirements (AT-0315).

OPEN
B33

Prepositioned infrastructure

Every arm with a net advantage relies on a medium already prepared along the whole route before departure. Once one-time preparation is charged separately, the advantage is pre-established remote state, not a shortcut (AT-0318, AT-0321).

OPEN
B40

Local-scale power ceiling

Under bounded amplitude, bounded rate, a gradient bound and enforced restoration, the best gated apparent/local leverage found in the sampled family is about 1.5 (AT-0362).

OPEN
B41

Leverage is action-expensive

Accounted action grows superlinearly with leverage across the gated set; the top-leverage case costs a large toy action. No cheap regime found (AT-0367).

OPEN
B46

Marginal cost is not cost

After preload each trip draws only 4.40 units of fresh external input while the fully amortized cost is 49.40. No figure from the infrastructure branch may be quoted without preload, depletion, recharge and maintenance amortized into it (AT-0381, AT-0382).

OPEN
B47

Infrastructure energy scales with route length

Preload and per-transit depletion are linear in the number of edges at 6.00 units per edge, R² above 0.999, while the trigger stays flat. A small trigger covering a long route means a long route was charged in advance, and delivering that loading locally adds overhead and setup time (AT-0383, AT-0385).

OPEN
B48

Prepared geometry is route-specific

A prepared corridor serves only the endpoints it was built for, and a medium general enough for arbitrary endpoints costs with area rather than length. Prepared links change what a trip costs, not what distance is (AT-0386, AT-0389).

OPEN
B43

Phase-switch amplification is state debt

Enormous gain from a tiny trigger in the single-field bistable medium is bought by leaving the field in a persistent alternate state. Gain with a nonzero final field is spending, not control (AT-0371).

OPEN
B44

Restoration and propagation do not co-occur

Adding a recovery field restores the medium to 1e-8 … 1e-4 but confines the response to a few cells. Across the sampled parameter plane no configuration both propagated long-range and restored (AT-0372, AT-0373).

OPEN
B45

Long-range propagation is bought from a filled reservoir

With explicit per-cell free energy, reach tracks stored energy and preloaded infrastructure grows linearly with route length. Labelled PREPOSITIONED INFRASTRUCTURE, never free propagation (AT-0374, AT-0375, AT-0376).

OPEN
B42

Consistency is not power

Multi-observer agreement and clean causal ordering are hygiene requirements the branch meets. They carry no evidential weight and are never reported as progress (AT-0361, AT-0368).

OPEN
B34

Damage rides with the corridor

The traveller's internal state is driven by the compression gradient it sits in, so the corridor that carries it is also what disturbs it. At matched progress the ride peaks about twice as high as the walk (AT-0331, AT-0333).

OPEN
B35

Smooth or affordable, not both

Peak identity disturbance falls with corridor width while accounted action rises with it. The affordable corridors are exactly the sharp ones (AT-0334).

OPEN
B36

Round trips accumulate identity debt

The identity drive depends on the magnitude of the compression gradient, not its sign, so the return leg adds to the residue rather than cancelling it (AT-0338).

OPEN

Full ledger: /blockers.

Most important unresolved question

Everything below is now downstream of one question. If emergent geometry comes from a deeper state space through a fixed, tame map, closeness underneath bounds closeness above and no mechanism written on top of that map can escape it. So: which of the four escape conditions — non-Lipschitz or critical regions, a dynamical map or metric, a lawful path metric that differs from the static state metric, or a quotient / branching / topology change — is mathematically NECESSARY for the core objective, and can it arise from the SAME underlying equations that already reproduce ordinary finite-speed behaviour, with nothing inserted by hand? Until that is posed as a theorem, no new mechanism family is opened. The older fronts remain live but subordinate. Four unresolved fronts now run in parallel. In the single-layer track: find a frozen, transferable, local, bounded rule that lowers the peak identity-damage floor at matched relational progress and action — without suppressing the interaction, without redefining identity vacuously, and without leaving state debt behind. In the two-layer track: exhibit a projection class and a lawful Y path whose normalised leverage genuinely exceeds what a matched random map supplies, whose projected trajectory does not sweep the intermediate region, and whose fast channel comes with a consistent deeper ordering. In the dynamic-geometry track: find a price, forced by the model rather than declared, at which editing the relationship map beats traversal in total accounted action while the edit itself does not have to sweep the intervening region. In the corridor track: derive the geometry-control price instead of assuming it, couple identity through the crossing, and show that a bounded corridor can beat ordinary traversal in total accounted action without depending on perfect synchronisation.

Next three experiments

AT-0407

Necessary escape condition

Which of E1 to E4 is mathematically NECESSARY for the core objective? Prove that the objective implies the failure of at least one named hypothesis, rather than assuming it (B53).

AT-0408

Escape from the same equations

Can the necessary escape condition arise from the SAME underlying equations that already reproduce ordinary finite-speed emergent behaviour, with no second channel inserted by hand (B54)?

AT-0409

Pricing topology change

E4 is the only unpriced door. Can a quotient, branching or topology change be given a local, bounded, restorable cost inside a toy model, so that it becomes auditable rather than merely unmodelled (B53)?

AT-0341

Coupling-law generality

Is the gradient-to-identity coupling forced in any model where geometry is dynamical, or can a lawful coupling be written in which a moving compression leaves internal state untouched (B34)?

AT-0342

Front-precedence theorem

Prove or refute: for every local control field with a bounded edit rate, the control front must reach the destination before the traveller can gain any time. AT-0332 supports it for one family only (B28).

AT-0343

Derived geometry price

Can a cost for editing a link be derived from an action principle inside the model rather than declared, and does any derived price change the verdict (B25, AT-0337)?

AT-0344

Carried traveller on a lattice

Repeat AT-0331 in two dimensions and under coarse-graining. G13 of the Carried-Traveller Gate is NOT MODELED (B8, B27).

AT-0345

Sign-sensitive identity coupling

Is there a lawful identity observable whose drive is sign-sensitive in the gradient, so that a closed cycle can return the traveller without residue (B36, B11)?

How to start next session

  1. 01Start at /lab/foundation and rerun the Phase 47 package at seed 373. Reproduce the recorded row — 9,600 pairs in dimensions 3, 5, 8, 12, largest ratio 0.99982, zero violations, best blind leverage 0.962, gate 5 of 10 — before trusting anything downstream.
  2. 02The bound is the frame for the whole programme. Do not open a new mechanism family. Any proposal must first be identified as escape class E1, E2, E3 or E4, and must arrive with its bill (B52, B53).
  3. 03Never quote an unnormalised stretch or leverage figure. Normalise the Lipschitz constant first; the flattering column is meaningless on its own (AT-0404).
  4. 04Open /lab/resonance and rerun the Phase 46 null at seed 363 before quoting any adjacency statistic, and run the ablate-and-shuffle audit with it (B49).
  5. 05Open /lab/infrastructure and rerun the Phase 45 package at seed 353. Reproduce the recorded lifecycle — 360 units preloaded, 33.6 recharged, 1.6 in triggers, eight completed trips, marginal 4.40 per trip against fully amortized 49.40 and ordinary traversal 48.00 — before trusting anything downstream.
  6. 06Never quote a marginal per-trip figure anywhere in this project without the fully amortized figure beside it (B46).
  7. 07Open /lab/stored-medium and rerun the Phase 44 stored-medium package at seed 343. Reproduce the recorded rows — several-hundred-fold gain with a persistent final field in the single-field medium, end debt of 1e-8 … 1e-4 with a front of only a few cells in the two-field medium, and a balanced energy ledger with zero violations — before trusting anything downstream.
  8. 08Never quote an amplification number without the reservoir debit beside it (B43, B45), and never mix the three accounting models of AT-0379.
  9. 09Open /lab/scale-power and rerun the Phase 43 package at seed 343. Reproduce the recorded row — leverage about 1.5 at accounted action of a few hundred toy units, endpoint debt under 0.02, |dλ/dt| inside the bound of 3 — before trusting anything downstream.
  10. 10Never quote observer agreement as a result. It is hygiene (B42). The result is the ceiling: weak leverage bought at superlinear action (B40, B41).
  11. 11Always report the coupled-backreaction number, never the free-ride control, and keep the normalised null control on the same chart.
  12. 12Open /lab/pulse and rerun the Phase 38 package at seed 303. Reproduce the recorded net-time row — traveller arc about 13.3 against a baseline of 24, net completion about 43.2 against ordinary traversal about 24.0 — before trusting anything downstream.
  13. 13Never quote the traveller's shortened arc on its own. The honest number is net completion including front propagation and restoration, and it is currently 1.8x slower than walking (B28).
  14. 14Treat the standing medium as infrastructure with a break-even crossing count, never as spontaneous travel. It wins on amortized action after about nine crossings and never wins on time (B30).
  15. 15Open /lab/corridor and rerun the Phase 37 package at seed 293. Reproduce the recorded benchmark row (Gaussian corridor action about 36.4 at shrink 0.449, global rewrite about 101) before trusting anything downstream.
  16. 16Quote the corridor advantage only as COMPUTATIONAL RESULT / MODEL-DEPENDENT. It is a statement about an assumed cost rule, never about energy or physical space.
  17. 17The two open gate items are the work: identity is not yet coupled through the corridor (G10, NOT MODELED), and the full cycle still costs more than simply covering the distance (G14).
  18. 18Open /lab/geometry and rerun the Phase 35 dynamic-geometry package at seed 273. The continuity no-go is the frame for everything that follows it.
  19. 19Never claim a skip from a fixed map. AT-0273 closes that route as a model theorem; only a dynamical projection or metric remains open, and it must be charged.
  20. 20Score the dynamic arm on total action across both sectors and on whether the geometry edit itself had to cross the region (B20, B21), not on the traveller's arc alone.
  21. 21Open /lab/deeper and rerun the Phase 34 two-layer package at seed 263. Reproduce the random-map null benchmark before trusting any leverage number.
  22. 22Never quote raw projection leverage. It is EXPECTED BY MAP GEOMETRY until normalised and compared against a matched random map.
  23. 23The one positive result to protect is AT-0268: finite emergent propagation recovered from local Y interactions with no speed inserted. Any new model must keep it.
  24. 24Open /lab/schedules and rerun the Phase 33 package at seed 255 to reproduce the recorded end-of-session numbers before changing anything.
  25. 25Read the end-of-day gate on that page. The status to beat is NONE PASSED; note which requirements currently fail.
  26. 26Treat the peak-damage floor (AT-0258, blocker B15) as the primary target. Every new candidate is scored against it.
  27. 27Do not optimise endpoint identity. The score is peak_t |ΔI| at matched destination error, with residue read after coupling is restored.
  28. 28Any new mechanism must be frozen before evaluation on unseen systems. Per-case retuning is recorded as retuning debt, not as progress.
  29. 29Preserve all rejected history. Killed shortcuts stay on the site with their evidence, and no result is relabelled upward without new experiments.