Skip to content

Historical transport source reviews and Volts comparison

This record preserves the dated source searches and completed Volts exploration. Its apparatus proposals and next-step language describe their original context; they are not a live queue. The maintained transport protocol owns the mathematical measurement/error contract, and the execution plan owns any reopening. Frozen data, declarations and their hashes are unchanged.

Passive transport measurement protocol

Protocol version: draft 6; scheduling references reviewed 2026-09-27. Source reviews and frozen experiments retain their original dates and scopes. Admission: not_admitted. Physical protocol result: not_tested; no admitted measurement contract. Role: supporting P2 measurement annex. The primary programme studies generative coherence patterns and their possible physical interpretation; MOSTR intake and additional data searches are deferred. The execution plan owns resumption. The plan's three-part variable contract separates mathematical definition, physical identification and prospective evolution. This annex supplies measurement evidence; it does not redefine the variables or turn calibration into independent physical identification. Historical exploratory record: the 2026-09-18 Volts comparison concerns the earlier fixed-reference pure-EPI model and its nominal measurement map. Its affine control has lower error within that comparison; it is not an evaluation of the revised generative research programme. See the model boundary. Physical admission remains open.

Scheduling: no acquisition task is active here. The 2026-09-21 source review supported the first TCLab approximation, now complete with physical admission still open. The sole plan determines any resumption; there is no automatic Volts refit or broad data search. Available resources: computer only, as confirmed by the user. The route is public terrestrial data. Building/acquiring a laboratory setup is not an active dependency; the apparatus below remains a reference design. The equal-capacity P2 model and a separately named fixed-reference variant are implemented through one shared rational log/exp owner. Finding a scalar dataset does not admit its physical mapping. The fixed-reference variant also supports the explicitly exploratory, single-curve computation below.

This is the measurement annex for the plan's supporting P1-P5 bridge. Historical O/S programme labels do not schedule current work. It specifies one finite, falsifiable laboratory mapping of the canonical EPI channel. A successful prediction would support that mapping within its measured domain; it would not establish the origin of physical reality or a unique explanation relative to other models with the same observable equation.

Primary-source inventory

The bounded review used four search queries and inspected primary record pages and JSON metadata on 2026-09-17. No signal archive was downloaded, no signal was scored, and no outcome was used to choose a model. Publication, record-update and acquisition dates are different; an unknown acquisition date remains unknown.

Candidate and primary source License, access and dates Structure and channels Clock, calibration and holdout gate Status and reason
Experimental networks of nonlinear oscillators, DOI 10.5281/zenodo.3521009 CC-BY-4.0; open. Published 2019-10-28; record modified 2020-01-24. Acquisition dates not established. 28 electronic Rössler oscillators, 20 wiring configurations, one recorded state variable per node; 30,000 samples, 101 coupling settings, three repetitions. Inspected metadata does not establish timestamp spacing, sensor calibration, complete hidden state, or an independent capacity/clock bridge. Whole repetitions could be reserved only after these gates and file identities are resolved. not_admitted. Useful topology inventory; active multivariable dynamics cannot be assumed to be passive EPI diffusion. Coupling-setting labels are not calibrated canonical weights.
Switched capacitor filter transient responses for fault detection, DOI 10.5281/zenodo.11122666 CC-BY-4.0 declared; files restricted. Published 2024-05-06; created/modified 2024-05-08. Acquisition dates not established. Measured and simulated responses of second-order filters, with underdamped and overdamped cases. Exact wiring/channel map unavailable in inspected metadata. Nominal training responses and labeled nominal/fault test responses are declared. Sampling clock, switching input, sensor calibration and split independence require documentation. Simulations must stay distinct from measurements; fault labels cannot enter calibration. not_admitted. Access and causal/model specification unresolved; a driven switched filter is not the proposed isolated P2 transport experiment.

License and access statements were checked against the corresponding 3521009 metadata endpoint and 11122666 metadata endpoint. A declared license does not grant access to restricted files. Access was not requested and no account action was taken.

The earlier local inspection of 3521009 retained metadata and only its 9,991-byte Structure.zip: twenty connected 28-node, 42-edge graphs with the same labeled degree vector. It downloaded zero time-series files. Its local evidence is artifacts/research/zenodo_3521009_admission/inspection.json; that ignored artifact is not a portable public evidence bundle. The primary record advertises about 20.6 GB of files; this protocol does not authorize downloading that collection.

No suitable simple passive RC or thermal time-series source was verified within this search budget. The previously cited EPFL-hosted RC diffusion article could not be retrieved during this review. It is a pending apparatus lead, not an admitted data source, license determination or checked calibration.

The second bounded review used eight targeted electrical/thermal queries and six primary-page inspections or retrieval attempts. It downloaded no signals, opened no reserved response and admitted no additional dataset. A failed page retrieval is an access limitation, not a finding about an unseen dataset.

Primary source Verified access or size Reason it does not supply this protocol
Supercapacitor discharge measurements, metadata Open CC-BY-4.0; ZIP 78,858,190 bytes; published 2026-03-25 Active constant-current electronic load, not isolated charge sharing between two measured storage nodes; full clock/calibration contract not established
INRIM 1 F supercapacitor measurements, metadata Open CC-BY-4.0; 21 files, 274,853,203 bytes; published 2026-01-23 One device with electrochemical test protocols; no verified simultaneous passive pair. The 270,585,991-byte training file alone exceeds the proposed download cap
PhysLab capacitor apparatus Public teaching page; open-data license and raw file sizes not verified One capacitor/resistor arrangement, not two observed storage coordinates; raw independent runs and instrument uncertainty not identified
USGS thermal equilibration record, official catalog Catalog declares public access; DOI retrieval failed; file sizes/license not verified Bath-driven biological temperatures; isolated two-body/common-capacity assumptions unmet in the catalog description
Han Hu Lab dataset inventory Search-index description advertises thermocouple measurements; direct retrieval returned HTTP 502 Isolation, exact files, license, calibration and repeat identities remain unverified; not an admitted source
Dryad thermal electrosurgery record Public download links; advertised archive 273.22 MB; published 2024-08-16; license not verified in inspected content Driven tissue heating and additional dissipation; incompatible passive P2 scope and oversized pilot archive

These exclusions concern the chosen experiment. They do not establish that other terrestrial TNFR mappings are impossible. Expanding the pressure law to use a convenient dataset would require a separately derived and admitted model; no such expansion is made here.

Multinode source review — 2026-09-18

A bounded follow-up used four targeted queries and four primary-page inspections/retrieval attempts. The selection question was whether a source could support a frozen structural prediction across preparations or a declared observation reduction, rather than another isolated scalar curve. No signal files were downloaded or response plots/tables intentionally inspected. Search snippets exposed published qualitative summaries; those summaries were not used to choose a model or evaluation window.

New lead Verified metadata Admission gap
Adelaide RC-ladder/PicoScope project University project proposes measuring several RC ladder circuits with a PicoScope No downloadable measurements, completed results, explicit circuit/channel map, uncertainty specification or independent preparation identities in the inspected page
ETH BMHT RC-Ladder System Identification, official inventory Inventory describes analytical identification code, AGPL-3, linked to DOI 10.1109/TCSII.2023.3340505 Direct repository retrieval failed; measured files, passive intervals, data license and independent preparations remain unverified. A code license is not a measurement-data contract
Kiel spatial soil-temperature archive, authors' data article Thermocouple time series and fixed sensor positions under controlled steady/cyclic loading around an electrical heater Driven spatial transport is outside the current isolated pure-EPI model. Sensor positions do not specify a sufficient finite nodal state. Archive retrieval failed; exact files, sizes, calibration and independent preparations unverified

No new candidate is admitted by this review. A failed retrieval gives no finding about unseen file contents. A future multinode protocol must retain enough initial-state information for prediction; hiding an indispensable sensor is not a valid test of emergent dynamics. This inventory starts no additional experiment or pressure-law extension.

Renewed terrestrial source review — 2026-09-21

The user explicitly requested real-world examples after reviewing the RC proposal. This pass distinguishes access to useful measurements from admission of a physical TNFR model. No fully documented passive RC network was admitted. Driven thermal and active electrical data are useful alternatives for specific questions; they cannot silently inherit the isolated two-node contract below. No model was fitted or scored, no signal archive was downloaded, and no reserved temperature/voltage arrays were inspected. Two small TCLab example CSVs were inspected only for schema and timestamps. Published qualitative summaries were seen; they are prior exposure, not blind evaluation evidence.

Source and access Verified observations Useful question and admission boundary
TCLab / Pyomo.DoE, authors' apparatus/data description Two heater/sensor assemblies; recorded seconds, two temperatures in Celsius and two heater commands in percent. Step/sine measurements and twelve repeat-labelled files are published. CSVs are about 20–29 kB. Repository BSD-3-Clause license; no separate data license found. First admission target: a reserved temperature response to a specified input, with partial-state/memory analysis. Heating, ambient losses and latent heater temperatures require an explicit driven model. Percent command is not independently measured watts. Per-run acquisition provenance, resets, calibration, ambient conditions and uncertainty remain open.
NED3-008: six thermocouples in a conduction apparatus, metadata CC0; latest advertised archive 253,223,992 bytes. Nine graphite and four titanium samples, with/without interface material, three labelled runs; seconds and six Celsius columns. Includes CAD and uncertainty notebooks. Spatial observation and boundary/contact effects. Intended for steady-state conductivity; transient applicability is unverified. No heater-input column in the stated schema. README's 1000 Hz claim differs from the named NI-9210's 14 S/s aggregate specification. Resolve logging versus fresh sensor conversions before defining a dynamical clock. Six sensors do not automatically define six closed thermal nodes.
28 electronic Rössler oscillators, primary apparatus article CC BY 4.0; twenty known graphs, 101 coupling settings and three repeated scans. The article now supplies nominal 30 kS/s acquisition and third-order 1500 Hz Butterworth filtering. Smallest signal archive: R2.7z, 1,027,983,572 bytes; topology archive only 9,991 bytes. Phase, synchronization and hidden-state tests. One of three circuit variables is observed; active terms remain in its equation. Instrument accuracy, channel skew and filter treatment remain unresolved. Reserve whole scans, not dependent files within one scan. The article's zero-coupling filename conflicts with its indexing rule; retain that ambiguity. Large signal downloads remain deferred.
Nonlinear analog learning network, 2024 data, metadata CC0; advertised data archive 15,756,671 bytes. The README exposes 4×4 node states under free/clamped conditions, edge gate-capacitor voltages and learning intervals; separate oscilloscope data have seconds and input/output volts. How node response depends on changing connections and supplied training. This is an active, externally tasked learning system, not autonomous TNFR formation. Measurement-step snapshots are not automatically continuous full-state trajectories. Gate-voltage-to-conductance mapping, full timing, calibration and run independence need admission. Archive retrieval was not verified.

Resolved prior lead. ETH's RC-ladder identification repository can now be read through its public GitLab API. The author paper, section III, uses 1,900 synthetic models; the inspected repository supplies code rather than measured trajectories. Close the previous access blocker as a negative measured-data lead. It remains mathematical identification work, not a physical RC dataset. Other circuit descriptions without published voltage measurements do not supply the missing observations.

Pinned TCLab intake and leakage boundary. Metadata was checked at commit d250c5b0625afb35007c075df1e0f7125e74017d. The example files each contain 901 timestamps spanning about 900 seconds, with approximately one-second spacing; this is a nominal recorded cadence, not a clock-accuracy certificate. Git blob identities expose two aliases:

  • tclab_step_test.csv and validation_experiment_env_2_step_50_run_1.csv: 3af095db6f6d13f6b80d0ab0c8cc4a57ca3928de.
  • tclab_sine_test_5min_period.csv and validation_experiment_env_2_sin_5_50_run_1.csv: 59a32387a0fd15d120df06531fefd84a1afdb373.

They are the same evidence under different names and must never cross a calibration/evaluation split. The twelve env_2 filenames group four input conditions by run_1, run_2, run_3; names alone do not prove independent preparations. Do not interpret period/amplitude labels without their owner specification. Published example outcomes are exposed; retain them as development evidence rather than claim they are unseen.

Follow-up completed under exploratory scope. The user authorized a first approximation while physical prerequisites remained unresolved. The TCLab experiment owner retains its declared nodal/input model, acquisition caveats, frozen split, results and code. It uses a separate driven affine realization; the equal-capacity P2 scorer was not relabeled as a general thermal predictor. No author-fitted parameters or thermal law were promoted to a TNFR derivation. The remaining sensor/clock, ambient, input and initialization gaps still block physical admission. Selection of this intake was based on accessible inputs, outputs and repeated-file structure, not observed TNFR prediction error. The execution plan owns scheduling; the other candidates remain alternatives, not parallel experiments.

Public metadata snapshots, file sizes/identities and review provenance are retained locally in artifacts/research/terrestrial_source_review_2026_09_21/. These ignored artifacts aid resumption but do not replace the cited public owners or constitute an independently timestamped prediction. P2 remains not_admitted; the source review itself supplied no performance result. The later exploratory scores belong only to the linked TCLab experiment.

Candidate-specific follow-up, 2026-09-26. The regional phase/amplitude contract rechecks these four sources against the derived directed six-node law. None is admitted for that prediction. That annex owns the new observation and resolution requirements; this inventory and the frozen TCLab model are not silently converted into a directed-diffusion experiment.

Bounded directed-source decision — 2026-09-26

The regional measurement contract motivated one bounded review of six new candidate studies, with four targeted queries each for electrical, thermal and material transport sources. An article and its linked data/methods records count as one candidate. This completes the planned search; it does not start a recurring discovery task. Only primary descriptions, methods and file metadata were assessed. Published summaries and figure captions were exposed, but no signal workbook/archive, response array or plot image was opened, no response scored and no model fitted. Selecting a lead on that basis is not externally blind validation.

Primary study / public record Observations and access established Decision for the fixed directed six-node prediction
Spatiotemporal diffusion metamaterial, 2020 Electrical apparatus with 50 modulated capacitors and 49 switches, maintained boundary voltage and time-averaged observations. Article CC BY 4.0; data statement points to paper/SI, with no linked machine-readable trajectory archive established. Not admitted: changing storage/support and homogenization differ from the fixed generator. Independent preparation, averaging and simultaneous-state evidence are unresolved.
Effective anyonic Bloch-oscillation circuit, 2022 Measured apparatus is an LC network of 23x23 node pairs with harmonic excitation and repeated node captures. The RC realization is an additional design. Article CC BY 4.0; supporting data/code require author request. Not admitted: additional state, forcing, sequential captures and Hilbert phase do not supply the required scalar observation/law; no verified public raw archive.
Electronically driven thermal pulses, 2026 Ten temperature nodes with programmed thermoelectric coupling; source XLSX advertised. Article CC BY-NC-ND 4.0; separate data-reuse terms, workbook size and independent run identities unverified. Not admitted: a different actively controlled chain. Its reported linear settings have a negative neighbor coefficient, outside the nonnegative pure-EPI realization; see the exact sign check below.
Temporal anti-parity-time transport, published 2025 / issue 2026 Rotating thermal architecture with timed changes in convection/material control and advertised figure-data XLSX files. Independent complete-state acquisition schema, clock uncertainty and data license not established. Not admitted: time-programmed distributed dynamics require their own observation/reduction law. Source-data availability alone does not establish a closed fixed six-node generator.
MOSTR apparatus, 2026, versioned dataset Three stirred tanks in series with outlet tracer measurements. Public CC BY 4.0 workbooks include a triple-system experiment and separate conductivity/flow calibrations. Not the triangle model. Selected only for a bounded alternative-graph admission check, owing to known one-way connectivity, accessible measurements and calibration files. Physical admission remains open.
Stirred-tank particle tracking, 2026 preprint, experimental-data DOI Particle trajectories in a laboratory vessel; compartment networks require an additional inference. Data resolver returned 404 in this review. Analysis-code record does not establish experimental-data access or its license. Not admitted: particle positions are not independent scalar compartment-state observations. Deriving the graph from the evaluated trajectories would need a separate calibration/reduction and holdout contract.

Constitutive sign check, not a new pressure implementation. The thermal pulse study declares a linear row, in physical seconds,

\[ \dot T_i=(C_d+C_v)(T_{i-1}-T_i)+(C_d-C_v)(T_{i+1}-T_i),\qquad C_d>0. \]

Applying the existing maximum-principle criterion to these observed nodes in a common positive affine temperature chart gives nonnegative pure-EPI weights if and only if abs(C_v)<=C_d. Their sum is 2*C_d, so with structural time t=s/tau_*, capacity would be nu=2*tau_*C_d. The reported linear pairs (0.006,0.010) and (0.003,0.053) instead give negative right-neighbor coefficients -0.004 and -0.050 per second. Positive capacity rescaling cannot remove that sign; clipping would change the apparatus law. This rules out that specific mapping, not physical feasibility, all other TNFR pressure laws or arbitrary hidden-state reductions. No signed-conductance exception is added to the engine.

Decision: park the exact six-node physical test at this source boundary. Its mathematical result, observation contract and uncertainty criterion remain valid under their hypotheses. No evidence here falsifies every TNFR model or proves that a suitable public dataset does not exist. The selected alternative below tests prospective nodal transport, not the triangle phase-locking theorem.

MOSTR: selected intake, not physical admission

The dataset, DOI 10.17632/z24ggxmg5f.3, was published on 2026-02-19. Its experiment metadata and calibration metadata advertise these files; their bytes have not been downloaded or checked:

File Advertised bytes Version-3 file identity
MOSTR_Triple_System_Residence_Data.xlsx 428,951 e399c85d-e0a5-4234-9ab2-f8c374e55637
MOSTR_Conductivity_Probe_Calibration.xlsx 50,786 672a6653-dc05-4c67-9572-0dd4d2ba29ba
MOSTR_Flowrate_Calibration_Example.xlsx 91,764 a7ecc20e-b92c-4983-9dae-3f0e043cce7b

The first file's advertised SHA-256 is a3bfc67847879452f5f6602397befd2ff954e6b2e52b4d24af3e95381082f860. This is repository metadata, not local byte verification or a forecast seal.

The author manuscript describes the triple experiment with balanced flow 0.175 L/min, hydraulic head 20 cm and tracer injected at the first vessel's inlet. Its generic procedure separately describes a 20 mL injection through a vessel lid; those descriptions do not establish the same event geometry or concentration waveform. The paper describes five-second IoT averages; the triple workbook's actual windows remain unverified. Example/representative calibrations are not automatically the actual three sensors' calibrations. The teaching-session record lacks hydraulic heads; it is not substituted for the triple workbook.

Conditional model, an inference to check rather than an apparatus finding. For fixed positive effective volumes V_i, balanced through-flow Q, ideal mixing and independently specified inlet concentration c_0(s), conservation of tracer would give

\[ V_i\frac{dc_i}{ds}=Q(c_{i-1}-c_i),\qquad x_i=(c_i-c_{\rm ref})/c_*,\qquad p_i=x_{i-1}-x_i,\quad \nu_i=\tau_*Q/V_i. \]

Here c_*>0 and t=s/tau_*. The existing directed pure-EPI row reads one upstream neighbor, opposite the material-flow arrow under engine conventions. This is a proposed physical bridge to an existing model, not a derivation of mixing physics from the nodal product. Its inlet and drain are supplied boundaries. An unresolved injection is an unresolved input event; it must not be inferred from the reserved response or silently replaced by an exact impulse or desired initial condition. A post-injection forecast instead needs an independently justified cutoff, initial state and later upstream baseline.

For constant inlet, deviations have a triangular generator with real diagonal eigenvalues -Q/V_i in physical time s, or -tau_*Q/V_i in structural time. Repeated rates permit polynomial-times-exponential terms. The directed dynamics owner supplies the relevant nonnormal scope. The two-triangle complex modes, phase lag and frozen-ratio discriminator do not transfer to this cascade. Its question is whether independently specified local transport predicts downstream response with the same observation model and no response-dependent pressure.

Before prediction, the schema must establish sensor-to-vessel identities and calibration applicability, conductivity-to-concentration/temperature mapping, effective volumes, balanced flow, injection timing, initial-state information, averaging windows/timestamp alignment and independent preparation identities. Bypass, stagnant regions, sensor lag and incomplete mixing are possible closure failures, not fitted corrections to add after evaluating a response. A first five-second average is not automatically the instantaneous initial state; a forecast must apply the declared averaging observation as well. If these dependencies or a separate evaluation acquisition are absent, retain the precise gap and do not claim physical P3 admission. The execution plan alone schedules that bounded schema check, now deferred under the collective-emergence priority. The frozen TCLab comparison remains unchanged.

Historical scalar-source selection — 2026-09-18

The decisions below belong to the completed Volts exploration. They are preserved for provenance, not instructions to reopen that intake or override the current source decision.

The focused scalar-data search retained two actionable leads. A single recorded curve is not two independent preparations.

Candidate Pinned observation metadata Remaining admission gaps
Stat2Data / Volts manual, package/license Package 2.0.0, GPL-3; 50 author-measured voltage/time observations, nominal 0.02-second spacing; one documented capacitor discharge after battery charging. Author-repository commit 3fe987c7e1813fd9960a3ca9293873359c1e07e9; data/Volts.rda, 767 bytes, Git blob SHA-1 6602a54be75afdec50b3a179457fc2af483bf29c No documented discharge wiring/source isolation, fixed-reference verification, voltmeter/clock accuracy, or independent run identities. Accessible exploratory candidate, not a reserved physical-validation dataset
Low-Energy-Meter, directory metadata no_load-f1Hz.csv, 614,400 bytes; Git blob SHA-1 c2ea34ca95407bdccb7e82f55b192f68094d4adb. README describes nanosecond timestamps, discharge-cycle IDs and ADC counts; the no-load file concerns leakage/instrument loading No explicit data license or verified independent repeat count, sensor/clock bounds or constant-reference evidence. README contains contradictory inverse versus proportional count-to-voltage formulas. Published README outcome summaries have been seen; this is not an untouched blind source

The Volts file is fixed at its exact revision. The historical admission preparation selected this 767-byte file for caching, with a 1,024-byte transfer cap and a Git-blob identity check before retention. The initial cache was not parsed, fitted or scored. That byte/provenance record is artifacts/research/stat2data_volts_admission/inspection.json; it records the actual raw-data SHA-256 without admitting a physical mapping. The retained 767 bytes have SHA-256 e50e505783dbf9b6d477165c92328e805e5d8bb0d27e3458862e2d5ba7f00a52. The initial cache delivery performed no conversion/decoding. Draft 4 later permitted the bounded exploratory decoding specified below. Nominal sample spacing is not clock accuracy.

The pinned package DESCRIPTION and the official CRAN index declare GPL-3. The generated package overview still says GPL-2; retain the pinned release declaration rather than silently mixing these metadata versions. The follow-up primary-source review did not resolve the apparatus, uncertainty or independent-preparation gaps.

Do not repair missing independent preparations by splitting this one curve and calling its windows independent runs. It can support a separately labeled exploratory study only after its narrower observation/model scope is declared. Physical P3 admission retains its missing calibration and independence gates.

Predeclared Volts exploratory delivery (before decoding values)

This is a single fixed computation inside P2.3, not physical P3 admission. The mathematical fixed-reference implementation is reusable; the following observation map remains a conditional hypothesis for this incompletely documented source. Stop after this one specification, including an unfavorable or unavailable result. No reference, split, horizon or error allowance is selected after looking at the curve.

  • Verify the pinned SHA-256 and 767-byte file before decoding. Bound expanded serialization to 65,536 bytes. Use a data-only R serialization decoder, not R evaluation. Require exactly one Volts data frame, 50 finite numeric rows, Voltage and Time columns, and increasing times. No downloads, gap interpolation, smoothing, row deletion or synthetic replacement in the run.
  • Read voltages as nominal scalar EPI coordinates (unit scale, zero offset) and reported seconds as nominal structural time (bridge one). These are stated chart/clock conventions, not sensor or physical-clock calibration. Set the candidate constant reference to zero before decoding; this is an unverified fixed-reference hypothesis, never a fitted asymptote.
  • Fit only the first and 25th samples (zero-based indices 0 and 24) using the continuous fixed-reference logarithmic identity. The first 25 rows are calibration; rows 25 through 49 are the continuation comparison. Interior calibration values may be used only by the declared comparison baseline, not to select another nodal model or rate after evaluation. Unresolved sign/decay or invalid times ends the specified nodal analysis.
  • Freeze the nominal rate enclosure and save a content-hashed continuation from the 25th value at the remaining reported timestamps before scoring continuation voltages. Reading that time schedule is declared metadata access, not evidence that these public samples were acquired prospectively. Predict through the existing rational exponential owner; no nodal Euler trajectory is invented or fitted to the continuation.
  • Compare coordinate RMSE/maximum absolute error and unchanged-initial-value persistence; include a calibration-only affine AR-1 baseline if its fit is identifiable, otherwise report baseline unavailability. These are descriptive controls, not additions to TNFR pressure. Save every specified result.
  • Instrument error, clock error, physical-reference adequacy and a physical acceptance threshold remain unknown, not zero. A rational enclosure of nominal arithmetic is not a measurement confidence interval. No numerical error, favorable baseline comparison or within-curve continuation closes the independent-preparation or physical-admission gates.

The saved protocol hash and result must retain within_single_acquisition, physical_status=not_admitted and measurement_verdict=not_assessed. The original byte-only inspection remains unchanged as historical evidence.

Completed exploratory record — 2026-09-18

Historical model boundary. This experiment predates the current foundation reassessment. It evaluated x'=nu*(0-x) with fixed positive nu, a zero-capacity reference node, pure EPI pressure and no active phase, capacity-contrast or topology source. Voltage was assigned directly to scalar EPI and reported seconds to structural time; neither mapping was independently calibrated. Those premises identify the tested model more precisely than a release label. The fixed-reference derivation remains a valid conditional reference; the experiment did not evaluate the complete engine or the current search for justified generative closure and identity formation.

Retain its data, frozen specification, source provenance and unfavorable comparison as historical evidence about that model/map pair. Do not transfer the outcome to a revised candidate without establishing equivalence of the tested observable prediction and measurement assumptions. Conversely, changing the programme or release name does not invalidate an unchanged prediction. A new candidate needs its own declared model boundary and independent reserved evaluation; these already inspected values cannot become fresh holdout data. Retiring the earlier candidate as the programme's representative does not provide the physical evidence still missing for the revised one.

The bounded reader and fixed-design benchmark executed the specification above without scientific changes. The archive expands to 1,016 bytes, R serialization version 2. Only one nominal rate was estimated: nu lies in the arithmetic interval [2.1590717641670265, 2.159071764167027] per reported second. This narrow interval encloses numerical arithmetic, not measurement uncertainty. The forecast starts at row 24 (t=0.48 s, V=3.2682 V); scoring uses only the 25 later observations through t=0.98 s.

Frozen predictor Continuation RMSE (V) Maximum absolute error (V)
Pure-EPI fixed-reference P2 0.0769495843 0.129818675
Persistence at the initial value 1.3676904144 2.028000000
Prefix-only affine AR-1 control 0.0255576220 0.046371615

The historical fixed-reference continuation improves on persistence but is less accurate than the declared affine control. It shows no predictive advantage for that specified model/map pair on this trace. It neither identifies the cause of the residual nor evaluates revised TNFR realizations: the observation map, constant reference, apparatus and error bounds remain unverified. No offset, rate, reference, split or pressure law was changed after scoring. Do not mine this curve for a replacement model and call the same continuation newly reserved.

The decoder initially stopped on historical unmarked text encoding, before issuing or scoring a forecast. A synthetic-tested strict ASCII metadata fallback repaired ingestion; it changed no measurements or scientific settings. Preserve this failed attempt alongside the successful execution. The whole archive was decoded before issue; the predictor received only prefix values and the reported future time schedule. This is explicitly within_single_acquisition, not blind prospective evidence. All physical statuses remain not_admitted / not_assessed.

Local retained bundle: artifacts/research/volts_exploration_2026_09_18/. Its execution manifest records base commit ce5d40feead702fc1a77fbea128ffa836a859081 with uncommitted source identified by sha256:f63bee1f1eeaed9e456224cf66530d483f8eebec2421f5ae3e093d8baddd9efa. The commit alone does not identify the executed source. Its predictor used the rational exponential reference, without a live graph trajectory. The exact pre-value protocol/specification snapshots are immutable:

  • protocol_before_values.md, SHA-256 e2b584003c2e5605ae786a533a48cb067a002e67f00a5ba9de2be8fc1f1c0124.
  • spec_before_values.json, SHA-256 b6a56c8856ea339778c6d6dfffd4f256a03f1fa2ce2a5e398aefb457afe41da0.
  • attempt_01.json: decoder stop; no forecast or score.
  • run_v2/: raw data, copied protocol/specification, ingestion metadata, issued forecast, complete residuals and byte-checked evidence envelope. Forecast content hash: sha256:57bb1e059470d80dbd6e685c6b5458857f702390018ccb31e080656f6102990c.

These are local ignored artifacts, not a published bundle or an independently authenticated timestamp. Python >=3.11 and pip install -e ".[research-data]" provide the pinned decoder. A replay uses the retained inputs and a fresh output directory; it cannot become a new blind study:

python -X utf8 benchmarks/volts_fixed_reference_exploration.py `
  --data artifacts/research/stat2data_volts_admission/Volts.rda `
  --spec artifacts/research/volts_exploration_2026_09_18/spec_before_values.json `
  --protocol artifacts/research/volts_exploration_2026_09_18/protocol_before_values.md `
  --expected-spec-hash b6a56c8856ea339778c6d6dfffd4f256a03f1fa2ce2a5e398aefb457afe41da0 `
  --output artifacts/research/volts_exploration_2026_09_18/replay

The combined focused suite passes 174 tests: both mathematical models, the unchanged equal-capacity acquisition contract, bounded ingestion, prefix isolation, baseline failures and complete evidence recording. The source-bound local checkpoint is artifacts/research/fixed_reference_volts_validation_2026_09_18.json. Tests validate these software contracts, not the physical hypotheses.