Skip to content
Adjacency Theory markAdjacency Theory

Phase 70 · physical mapping · first bench experiment · AT-0689 → AT-0700

We stop arguing about space and go build a small electronic network that can tell us we are wrong.

Everything the programme has built so far lives inside mathematics. This page is the design that leaves it: a programmable network of electronic oscillators with local, logged coupling updates, and ten pass/fail metrics locked before a single component is bought. The goal is not spacetime manipulation. It is to physically realize the abstract two-phase moving-doorway architecture and find out whether measured behaviour matches what was written down in advance. Nothing here has been built, run, or measured.

Status

physical evidence: none

Experiment v0.1 — pre-registered, unbuilt, and designed to fail loudly.

PROTOCOL ONLY — NO HARDWARE DATA EXISTS

Locked metrics

10

M1 → M10, fixed before any run

Controls

7

including an intentionally cheating global switch

Declared null

N = 8

no advantage expected or permitted

Physical evidence

NONE

Bench wiring schematic · stage 0 · 16 oscillators · degree 3

design targets · not measurements
0123456789101112131415
  • W1Sixteen oscillator nodes on a bounded-degree-3 graph: 24 physical links, each carrying one programmable coupling element. Physical degree is fixed in copper and never changes during a run.
  • W2Each coupling element exposes a single digital weight register. C0 is the low-weight baseline; C* is a transient high-weight state. The graph is the same in both phases — only weights move.
  • W3Every node exposes an analogue probe point to a synchronised ADC channel, plus a second channel for the coupling-element drive, giving 32 capture channels at Stage 0.
  • W4The controller is partitioned in firmware so that the logic for node i may read only node i's own state, its ≤3 neighbours, and an address prefix. This partition is the anti-cheat boundary and it is audited statically as well as from the log.
  • W5Four supply rails are physically separate from the start: control logic, oscillators, coupling elements, and the restoration drive. Energy is never apportioned by subtraction.
  • W6Fault injection sits in series with eight of the links and in the power gate of four nodes, driven from the same clock as the log.

Solid ring links and dashed chords are physical copper and never change. Only coupling WEIGHTS move between C0 and C*. The highlighted path is an illustrative doorway route, not a measured one.

Illustrative run timeline · trigger → moving window → arrival → recovery

physical evidence: none
  1. 1 · Trigger

    Local seed crosses threshold at node 0.

  2. 2 · Moving window

    A bounded active set advances along the address route.

  3. 3 · Payload arrival

    Four-component waveform captured at the destination.

  4. 4 · Recovery

    Couplings fall back to C0 behind the front. No global reset.

ANIMATION OF THE PROTOCOL'S EXPECTED SHAPE. No instrument produced this trace.

Preregistration scorecard · ten locked metrics

physical evidence: none
M1Trigger thresholdlocked · awaiting hardware

An intentional local seed crosses a reproducible activation threshold. Below-threshold perturbations launch no doorway above a false-trigger ceiling declared before the run.

fails if Response is smooth with no threshold, or below-threshold noise launches doorways above the ceiling.

Instrument — Amplitude sweep, 200+ trials per amplitude, operator blind to outcome.

M2Localized activationlocked · awaiting hardware

Maximum simultaneously active nodes stays at or below c · log2 N, with c fixed before the run; the initial target window is equivalent to about four nodes at small N.

fails if Active count grows like O(N) — a global avalanche rather than a moving doorway.

Instrument — Synchronised per-node ADC capture, active = amplitude above declared activation band.

M3Autonomous propagationlocked · awaiting hardware

The controller may implement only the same local neighbour update law at every node. No global path command, no destination-wide broadcast, no precomputed per-trip shortest path.

fails if Any coupling write depends on state outside the writing node's bounded neighbourhood and address prefix.

Instrument — Static audit of controller firmware plus timestamped control-write log.

M4Payload identitylocked · awaiting hardware

A structured waveform with at least four independent components, compared before and after transit against an identity-error tolerance predefined for each component.

fails if Any component exceeds tolerance, or a single tone is substituted for the structured payload.

Instrument — Multi-channel capture at source and destination, component-wise error metric.

M5Latencylocked · awaiting hardware

Total clock starts at trigger. C0 baseline and doorway mode are compared on identical hardware and topology. N = 8 is an explicit null where no advantage is expected; N = 16 / 32 / 64 test the trend.

fails if Doorway latency fails to fall below the C0 baseline at N = 16 and above, or an advantage is claimed at N = 8.

Instrument — Common time base, trigger-to-arrival timestamps on both modes, interleaved runs.

M6Restorationlocked · awaiting hardware

All coupling and state variables return within a declared baseline tolerance WITHOUT a global reset.

fails if Any global reset command appears in the log, or variables settle outside tolerance.

Instrument — Coupling-register dump and per-node state capture through the recovery window.

M7Energylocked · awaiting hardware

Supply energy measured separately for control, oscillators, coupling elements and restoration. Once hardware exists, no normalized proxy is permitted anywhere on this page.

fails if Energy is reported as a normalized count, or the four rails are lumped into one figure.

Instrument — Four instrumented supply rails, integrated current-sense over the whole transfer window.

M8Fault tolerancelocked · awaiting hardware

Edge and node failures are injected DURING transit; local repair and recovery are logged as they happen.

fails if Recovery requires a global rebuild, or the log cannot show the repair was locally decided.

Instrument — Switched link cut-outs and node power-gates on a scripted injection schedule.

M9Multi-flowlocked · awaiting hardware

Two, four and eight concurrent payloads at N ≥ 32, tracking p50 and p95 latency and every conflict event.

fails if p95 collapses, or conflicts are resolved by any central arbiter.

Instrument — Concurrent source triggers, per-flow timestamps, conflict counters in the log.

M10Anti-cheat auditlocked · awaiting hardware

Every control write is timestamped. Any nonlocal or global reconfiguration DISQUALIFIES the run from counting as doorway evidence.

fails if The log is incomplete, or a single global reconfiguration write is found.

Instrument — Append-only control-write log, audited independently of the narrative.

Anti-cheat control-write log

physical evidence: none
t (µs)writer nodelinkweightscope
000–10.050.82 openLOCAL
1211–90.050.84 openLOCAL
2400–10.820.05 closeLOCAL
3099–100.050.83 openLOCAL
4211–90.840.05 closeLOCAL
481010–110.050.81 openLOCAL
6099–100.830.05 closeLOCAL
721010–110.810.05 closeLOCAL

ILLUSTRATIVE LOG SHAPE — these rows show what an admissible log looks like, not what any instrument recorded. A single row stamped GLOBAL would disqualify the run.

Scaling design space · N = 8, 16, 32, 64

design targets · not measurements
N = 8NULL / CONTROLno advantage expected
ordinary chain
4
doorway target
4.8

At eight nodes the doorway overhead exceeds the ordinary chain. NO ADVANTAGE IS EXPECTED and none may be reported. This size exists to show the architecture does not win everywhere.

N = 16SCALING TESTtarget advantage 1.25×
ordinary chain
8
doorway target
6.4

First size where the design target predicts any advantage at all, and only a slim one.

N = 32SCALING TESTtarget advantage 2.00×
ordinary chain
16
doorway target
8

Stage 1 build. Bounded-degree graph, concurrency sweep begins here.

N = 64SCALING TESTtarget advantage 3.33×
ordinary chain
32
doorway target
9.6

Stage 2 build, physical or time-multiplexed. Time-multiplexed nodes are EFFECTIVE, not physical, which weakens any localization reading.

Ndegree-3 edgesmeasurement channelscontrol updates / ticktarget max active fractionordinary latencydoorway target
812161250.00%44.8
1624322425.00%86.4
3248644812.50%168.0
6496128966.25%329.6

These are EXPERIMENT DESIGN TARGETS and illustrative model expectations in arbitrary latency units. They are not measured hardware data and there is no measured column, because no measurement exists.

Controls · run interleaved in a pre-declared random order

idcontrolwhat it separates
K1C0 only — latent couplings disabledEstablishes the ordinary-route baseline on the same hardware.
K2Global switch control — intentionally cheatingEstablishes the upper bound of conventional switching performance, so a doorway result can be placed against it honestly.
K3Random local switching at matched control writes and energySeparates structured local law from mere reconfiguration activity.
K4Shuffled destination addressesDetects a route that works regardless of the address it claims to follow.
K5Below-threshold triggerMeasures the false-trigger floor that M1 is scored against.
K6Same payload over the ordinary routeSeparates payload fidelity from doorway transport.
K7Frozen doorway state, recovery suppressedIsolates the cost of restoration from the cost of transport.

Success classification · fixed before the run

A

HARDWARE ANALOG FAIL

No reproducible moving doorway. The architecture does not survive contact with a circuit and the toy result stands alone.

B

PROGRAMMABLE NETWORK EFFECT ONLY

An effect exists, but it requires global switching or path scripting. That is conventional network engineering, not a doorway.

C

LOCAL TWO-PHASE ANALOG CANDIDATE

Localized thresholded propagation, payload delivery and restoration under a purely local law, surviving every control. A statement about a circuit.

D

SCALING ANALOG CANDIDATE

Grade C, plus latency and active-footprint trends that continue favourably with N across the tested sizes only.

There is deliberately no grade above D. No outcome of this experiment can be classified as evidence about spacetime, distance, or fundamental physics, and no such category will be created.

Experiment v0.1 · bill-of-materials categories

categoryrolestage 0 qtyrequirement
Oscillator nodeOne tunable electronic oscillator per node — the substrate element S.16Stable free-running frequency, accessible phase/amplitude probe point, identical part across all nodes.
Programmable coupling elementSets the weight of each link between two nodes: C0 baseline low, C* transient high.24 (degree-3)Digitally settable coupling strength, monotonic response, settling time far below one propagation tick.
MCU / FPGA controllerRuns the identical local update law at each node and writes coupling registers.1 FPGA or 16 per-node MCUsDeterministic tick, per-node scope limited in firmware to its own bounded neighbourhood and address prefix.
ADC / scope channelsCaptures per-node state for the active-count, latency and payload-identity metrics.32 channelsSynchronised sampling on a common time base, bandwidth several times the oscillator frequency.
Power measurementFour separated supply rails: control, oscillators, coupling elements, restoration.4 instrumented railsCurrent-sense with integration over the transfer window; no rail may be inferred by subtraction.
Fault injectionCuts links and gates nodes during transit for the fault-tolerance metric.8 switched cut-outsScriptable, timestamped on the same clock as the control-write log.
Logging hostAppend-only capture of every control write, trigger, injection and capture frame.1Write-once storage, clock-synchronised, audited independently of whoever ran the experiment.

Component choices are GENERIC categories. No part number, vendor, price or availability has been sourced or verified, and none should be inferred from this table.

Data schema · append-only capture

run_metaonce per run

run_id · utc_start · N · topology_hash · firmware_hash · mode (C0 | doorway | control_id) · declared thresholds · operator_blind flag

control_writeevery write, append-only

t_ns · writer_node · target_link · old_weight · new_weight · justifying_neighbour_state · address_prefix_used

node_statesynchronised ADC frame rate

t_ns · node_id · amplitude · phase · active_flag (vs declared activation band)

payloadcapture window at both ends

t_ns · component_index (≥4) · amplitude · phase · source_or_destination

energycontinuous, per rail

t_ns · rail (control | oscillator | coupling | restoration) · instantaneous_power · integrated_joules

faultper injection

t_ns · injected_element · type (link_cut | node_gate) · restored_at_ns · repairing_nodes

flowper payload, per concurrency level

flow_id · trigger_t_ns · arrival_t_ns · conflicts · p50 · p95

verdictonce per run, written after the log is sealed

metric_id (M1…M10) · locked_criterion · observed · PASS | FAIL | NOT MODELLED · auditor

Step-by-step test protocol

  1. 01

    Seal the preregistration

    Publish c for M2, the false-trigger ceiling, the per-component identity tolerances, the baseline restoration tolerance and the control ordering. Hash the document. Nothing below may change these.

  2. 02

    Build and characterise C0

    Bring up 16 nodes at baseline coupling. Measure free-running frequencies, drift, noise floor and the ordinary chain latency across the declared endpoint pairs.

  3. 03

    Audit the firmware partition

    Statically verify that no node's update law reads outside its bounded neighbourhood and address prefix. A failure here stops the experiment; it is not a finding, it is a build defect.

  4. 04

    Threshold sweep (M1)

    Sweep trigger amplitude with 200+ blind trials per level, including the below-threshold control K5. Fit the formation probability curve.

  5. 05

    Single-payload doorway runs (M2, M4, M5, M6)

    At N = 8 first, as the declared null. Then 16. Capture active counts, four-component payload fidelity, trigger-to-arrival latency and the restoration tail.

  6. 06

    Interleaved controls (K1–K7)

    Run every control in the pre-declared random order, interleaved with doorway runs so drift cannot masquerade as an effect.

  7. 07

    Energy runs (M7)

    Repeat the transfer with all four rails integrated separately, including a recovery-suppressed K7 pass to isolate restoration cost.

  8. 08

    Fault injection during transit (M8)

    Cut links and gate nodes mid-flight on the scripted schedule. Log which nodes repaired and how far the repair propagated.

  9. 09

    Scale to 32, then 64 (M5 trend)

    Repeat steps 5–8. Report trends only; no exponent may be fitted from four sizes and none will be.

  10. 10

    Concurrency (M9)

    Two, four and eight simultaneous payloads at N ≥ 32. Record p50, p95 and every conflict event.

  11. 11

    Seal the log and audit (M10)

    Freeze the append-only log. An auditor who did not run the experiment checks every control write against the locality rule. One global reconfiguration disqualifies the run.

  12. 12

    Score and publish, pass or fail

    Fill the verdict stream against the sealed criteria, assign grade A–D, and publish the result unchanged. A failure is published with the same prominence as a pass.

For a ten-year-old

Now we stop asking the computer to pretend. We build a real electronic network and see whether a tiny moving doorway can appear for real. If the hardware disagrees, we change the theory — not the result.

Assembling the pre-registered protocol…