Adjacency Research Program · Build + Fund

Process validation

Validating the referee, not the result. Everything below tests whether our campaign logic can be fooled. None of it tests whether the physical effect exists.

simulation / process pipeline: INTERNALLY VALIDATEDphysical effect: NOT TESTEDcausal-bound anomaly: NOT TESTEDphysical evidence: NONE

← Build + fund the test

01 · Two different questions

answerable today

Validating the process

Asks: can this campaign, run exactly as frozen, reject worlds we know are fake and recognise a dataset we deliberately built to look like the target? This is answerable today, entirely in software, and we have answered it.

requires hardware

Validating the hypothesis

Asks: does the apparatus ever decode post-arming unpredictable information before the conservative lower bound of every inventoried physical path? This requires hardware that does not exist. It is untested and unclaimed.

02 · Synthetic end-to-end campaign sanity test — 1,000 campaigns per world

Synthetic world	How it was constructed	Candidate verdicts	Rejected by
W1 NULL	No mechanism at all. Transport is ordinary, timing is ordinary, nothing beats the causal bound.	0 / 1000	Latency and sparsity signatures never clear the frozen thresholds.
W2 HIDDEN-SHORTCUT	A concealed conventional bypass that genuinely beats the logical route — the most dangerous kind of fake, because the timing looks spectacular.	0 / 1000	Mechanism audit gate. The channel inventory finds the bypass and the entire campaign is invalidated before any timing number is considered.
W3 CONVENTIONAL SWITCHING	Ordinary fast switching plus adaptive caching — an evil twin that satisfies the latency ruler by design.	0 / 1000	Sparsity ruler and mechanism audit. Fast is not enough; the active fraction must also fall with N and the mechanism must be locally auditable.
W4 TARGET-LIKE	A dataset we intentionally constructed to satisfy every preregistered signature: falling latency ratio, falling active fraction, clean audit, clean quality gates.	1000 / 1000	Nothing — and it should not be. This world was built to pass.

What this tests

Whether the campaign logic rejects worlds we know to be fake.
Whether the audit gate outranks a spectacular timing result (it must, and it does).
Whether the pipeline retains enough sensitivity to recognise an intentionally constructed target-like dataset.
Whether any of this depends on a human choosing a window, a threshold, or a smoothing parameter after seeing the data (it does not).

What this does not test

Whether the physical effect exists.
Whether any apparatus behaves like the TARGET-LIKE world.
Whether our channel inventory is complete on real hardware — synthetic worlds only contain the channels we thought to put in them.
Whether real timing systems achieve the uncertainty these simulations assume.

03 · Stage-gate QA — B0 through B8

B0 · Clock loopback

Entry criteria: Both endpoints powered on separate domains; both timing stacks warmed and logged.
Procedure: Loopback timestamping over a characterised reference path, repeated across hours and temperature.
Raw outputs: Raw timestamp streams from both stacks, environmental log, loopback path characterisation.
Acceptance criteria: Both stacks agree within the stated combined uncertainty; drift bounded and documented.
Failure actions: Halt the programme. Rebuild the timing chain. No downstream phase may proceed on unverified clocks.
Re-test criteria: Full repeat from a cold start after any change to references, cabling or firmware.
Responsible role: Precision timing / metrology lead
Required equipment: 2x FPGA endpoints, timing references, independent time-interval instrument.
Configuration / firmware hash: PENDING — issued at firmware freeze
Calibration IDs: PENDING — instrument certificates required before entry
Evidence artifact: Measured timing uncertainty budget (signed, hashed).

B1 · Entropy and decode

Entry criteria: B0 passed and archived. TRNG installed with health monitoring.
Procedure: Arm, generate payload post-arming, transmit over an ordinary path, decode at B, hash payload at creation.
Raw outputs: Payload hashes, entropy health logs, decode records, arming timestamps.
Acceptance criteria: Decode success above the preregistered error threshold; payload provably generated after arming.
Failure actions: Fix the payload chain. Correlation without decode is never accepted as information transfer.
Re-test criteria: Repeat on any TRNG, encoder or decoder change.
Responsible role: FPGA / embedded lead
Required equipment: Hardware TRNG, encoder/decoder firmware, raw capture.
Configuration / firmware hash: PENDING
Calibration IDs: N/A — entropy health records instead
Evidence artifact: Validated information-transfer primitive with frozen decode rule.

B2 · Delay bank

Entry criteria: B1 passed. All delay segments installed and inspected.
Procedure: Bidirectional group-delay characterisation per segment, across temperature and time; OTDR on every fibre; spool-swap repeat.
Raw outputs: Per-segment delay tables with uncertainty, OTDR traces, thermal logs.
Acceptance criteria: Every segment characterised empirically with conservative bounds. Nominal length is never an input.
Failure actions: Re-characterise or replace the segment.
Re-test criteria: Recalibrate on any physical disturbance, reroute or calibration expiry.
Responsible role: Precision timing / metrology lead
Required equipment: Fibre spools, matched transceivers, OTDR, environmental sensing.
Configuration / firmware hash: PENDING
Calibration IDs: PENDING — one certificate ID per segment
Evidence artifact: Calibrated delay bank record.

B3 · Hidden-shortcut detection

Entry criteria: B2 passed. Injection fixture available; operator blinded to installed bypasses.
Procedure: A colleague installs concealed conventional bypasses. The inventory and audit pipeline must find every one before unblinding.
Raw outputs: Audit logs, RF and optical scans, detection record versus the sealed bypass list.
Acceptance criteria: 100% detection. No partial credit exists at this gate.
Failure actions: Hard stop. Extend the inventory and repeat until detection is total.
Re-test criteria: Repeat before every confirmatory campaign, not just once.
Responsible role: RF / EMI engineer with independent red team
Required equipment: Shielded enclosures, RF scanner, optical leakage detection, injection fixture.
Configuration / firmware hash: PENDING
Calibration IDs: PENDING — scanner calibration required
Evidence artifact: Detection record with sealed bypass list and reveal timestamp.

B4 · Baseline propagation

Entry criteria: B3 passed at 100%. Substrate assembled and inspected.
Procedure: Measure ordinary transport under the strongest permitted conventional signalling; build D_tau as the minimum over all inventoried paths.
Raw outputs: Per-path earliest-influence measurements, path inventory, session-to-session comparison.
Acceptance criteria: Reproducible baseline with a conservative lower-confidence bound.
Failure actions: Recharacterise the substrate. No comparison exists without a baseline.
Re-test criteria: Repeat on any topology or wiring change.
Responsible role: Principal experimental physicist
Required equipment: 16-node oscillator PCB, DAQ.
Configuration / firmware hash: PENDING
Calibration IDs: PENDING
Evidence artifact: Empirical causal envelope D_tau for the apparatus as built.

B5 · Analogue dynamics

Entry criteria: B4 passed and D_tau archived.
Procedure: Drive the substrate into the nonlinear moving-front regime; measure transport relative to C0.
Raw outputs: Raw traces, control logs, per-node activity records.
Acceptance criteria: Reproducible architecture effect relative to C0. Explicitly carries NO causal claim.
Failure actions: Record and publish the negative result.
Re-test criteria: Repeat across topologies without retuning.
Responsible role: Nonlinear dynamics expert
Required equipment: Substrate, control firmware, DAQ.
Configuration / firmware hash: PENDING
Calibration IDs: PENDING
Evidence artifact: Characterised analogue transport, bounded by D_tau.

B6 · Blinded mixed controls

Entry criteria: B5 characterised. Sealed condition schedule generated by a custodian outside the run team.
Procedure: Randomised mixed schedule of real, null, sham and bypass conditions; operators and analysts blind.
Raw outputs: Run records with condition IDs sealed, verdict assignments, unblinding log.
Acceptance criteria: Verdicts match conditions at the preregistered rate; no condition leaks through operator behaviour.
Failure actions: Redesign the run protocol. No confirmatory data may be collected under broken blinding.
Re-test criteria: Repeat after any protocol or staffing change.
Responsible role: Independent statistician / red team
Required equipment: Blinding harness, sealed schedule custody.
Configuration / firmware hash: PENDING
Calibration IDs: N/A
Evidence artifact: Blinded pipeline validation report.

B7 · Causal-bound confirmatory

Entry criteria: B0–B6 all passed, all hashes frozen, all calibrations current.
Procedure: Full delay-amplified campaign across multiple delay settings, with cable-swap nulls, topology permutation and dual timing.
Raw outputs: Complete raw dataset, both timing streams, environmental logs, all hashes.
Acceptance criteria: Delta positive by the preregistered guard band at every delay setting with all quality gates green.
Failure actions: Report NO ANOMALY or ARCHITECTURE EFFECT — the expected outcome, and fully publishable.
Re-test criteria: No re-runs after unblinding. A repeat is a new preregistered campaign.
Responsible role: Principal experimental physicist with external observer
Required equipment: Complete Golden Metrology Rig.
Configuration / firmware hash: PENDING
Calibration IDs: PENDING
Evidence artifact: Published dataset with hashes, whatever it says.

B8 · Independent rig

Entry criteria: B7 complete and published, including protocol and parts specification.
Procedure: A second rig built from the specification — not cloned — runs the same frozen campaign.
Raw outputs: Independent raw dataset with its own hashes and custody chain.
Acceptance criteria: Preregistered result reproduced with no retuning.
Failure actions: Report the failure to replicate as the finding it is.
Re-test criteria: External replication follows regardless of outcome.
Responsible role: Second team with facility partner
Required equipment: Full second apparatus.
Configuration / firmware hash: PENDING
Calibration IDs: PENDING
Evidence artifact: Replication dataset, or documented non-replication.

04 · Run manifest & chain of custody

Every run carries

01Run ID and UTC start/stop

02Protocol version + hash

03Firmware hash (both endpoints)

04Analysis-code hash

05Topology hash and wiring map revision

06Calibration certificate IDs, with expiry dates

07Delay-bank setting and characterised group delay

08Both timing stack identifiers and reference status

09Entropy source health record

10Payload hash issued at generation

11Environmental log reference

12Blind key custodian and seal state

13Operator and witness signatures

14Raw dataset checksum

mandatory stop conditions

Any calibration certificate has expired.
Any configuration or firmware hash differs from the frozen value.
The two timing systems disagree beyond the stated combined uncertainty.
Hidden-shortcut detection falls below 100% on any control.
Any raw-data integrity or checksum verification fails.

These are mandatory stops, not warnings. Data collection halts immediately and the affected runs are invalidated in full — never sample by sample.

05 · Validation matrix

Phase	Designed	Simulated / red-teamed	Hardware built	Calibrated	Blinded run	Replicated	External replication
B0 Clocks	●	●	○	○	○	○	○
B1 Entropy / decode	●	●	○	○	○	○	○
B2 Delay bank	●	◐	○	○	○	○	○
B3 Shortcut detection	●	●	○	○	○	○	○
B4 Baseline	●	●	○	○	○	○	○
B5 Analogue dynamics	●	◐	○	○	○	○	○
B6 Blinded controls	●	●	○	○	○	○	○
B7 Causal-bound confirmatory	●	◐	○	○	○	○	○
B8 Independent rig	◐	○	○	○	○	○	○

● complete◐ partial○ not started

06 · Evidence locker

Artifact slot	Format	State
Protocol version + hash	semver + SHA-256	issued at freeze
Firmware hash	SHA-256 per endpoint	issued at freeze
Wiring photograph + hash	Image + SHA-256	awaiting hardware
Calibration certificate IDs	Issuer, ID, expiry	awaiting hardware
Raw dataset checksum	SHA-256 over unsmoothed capture	awaiting hardware
Analysis-code hash	SHA-256, frozen pre-unblinding	issued at freeze
Blind key custodian	Named party outside the run team	custodian assigned
Result classification	Frozen verdict enum	awaiting hardware
Anomaly notes	Free text, append-only, timestamped	awaiting hardware

07 · Next 30 days

Finalise the partner technical brief

Two pages a physicist can dismantle in ten minutes.

Obtain timing, fibre and metrology quotes

Replace every REQUEST QUOTE line in the BOM with a real number.

Recruit experimental physics and timing/metrology advisors

Advisory first, employment later. Independence matters more than availability.

Identify a host lab

Facility with characterised fibre plant and calibration standards already in place.

Freeze B0–B3 wiring diagrams

Frozen and hashed before any procurement decision.

Prepare an NSF-style project summary and budget

Instrumentation framing, not physics-claim framing.

Secure the first $75k–$150k

Partner-assisted MVP envelope.

Procure P0 only after partner architecture review

No hardware is bought before an independent reviewer has attacked the design.