HONESTY.TOOLS

Some utilities are publicly released; other entries describe research source, bounded studies or methods still under development. Availability and scientific validation are separate questions. Try the model-free evidence workbench for invented interval, delivery and question-planning examples; it is a separate browser demonstration, not an extra catalog service.

Choose a research task

Instruments by research task

Availability and scientific validation are separate questions. Each instrument keeps its method, access and limits attached.

Inspect representations

Inspect representationsResearch prototypeJacobian bridgeHow does a small directional change affect a readout?Source not publicly releasedRead instrument details ↗
Method
Directional finite differences at selected model layers.
Output
Derivative/readout estimates and comparison outputs.
Limit
A local derivative estimate does not establish a general causal interpretation. Extraction and step-size contracts matter.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Checkpoint/residual access, layer/tap definition, prompt pool and finite-difference step contract.
Inputs
Prompts, selected residual directions/layers and step sizes.
Compatibility
Internal-access prototype; estimator and normalization must match the source.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-01 · Jacobian bridge · instrument:tool-01

Inspect representationsResearch prototypeRidge readout lensesWhat can a fitted linear readout recover at a layer?Source not publicly releasedRead instrument details ↗
Method
Fit and compare ridge readouts under a declared split and tap recipe.
Output
Readout performance and onset comparisons.
Limit
This is a fitted linear readout, not a Jacobian. Model and normalization identities must be recorded.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Checkpoint/residual access, normalization contract and explicit train/test split.
Inputs
Layer activations and corresponding prediction targets under a recorded split.
Compatibility
Fitted linear readout; not interchangeable with directional derivatives.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-02 · Ridge readout lenses · instrument:tool-02

Inspect representationsResearch prototypeSilhouette arbitrary-frame APIHow does a representation read under a researcher-defined frame?Source not publicly releasedRead instrument details ↗
Method
Compare embedding anchors defined by names and region descriptors.
Output
Frame rankings, margins and uncharted conditions.
Limit
A chosen descriptor frame is an instrument definition, not an intrinsic ontology or universal calibrated scale.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
An identified embedding encoder, descriptor frame and anchor/reference assets.
Inputs
Named regions/descriptors, embedding anchors and a query representation.
Compatibility
All frame and input embeddings must use the compatible encoder and preprocessing.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-03 · Silhouette arbitrary-frame API · instrument:tool-03

Inspect representationsIncomplete research sourceRotated lensWhat changes under a fitted within-model rotation?Source not publicly releasedRead instrument details ↗
Method
PCA, Procrustes alignment and linear-probe transport.
Output
Alignment and probe comparisons.
Limit
Recovered source has missing helpers, fit/score reuse and transport-algebra issues. It is not ready to offer as a validated instrument.
Maturity
Incomplete research source
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Missing source helpers must be resolved; compatible model activations and fit/evaluation inputs.
Inputs
Layer representations, paired examples and a declared probe/rotation recipe.
Compatibility
Incomplete source; fit/score reuse and transport algebra require review before use.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-05 · Rotated lens · instrument:tool-05

Inspect representationsIncomplete research sourceGrassmannian velocityHow do selected subspaces move across layers?Source not publicly releasedRead instrument details ↗
Method
SVD analogy subspaces, nuisance projection and principal angles.
Output
Consecutive-layer subspace distances.
Limit
Helper and result gaps remain. A subspace comparison does not establish a shared causal mechanism.
Maturity
Incomplete research source
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Missing helpers/results must be resolved; layer representations and nuisance directions.
Inputs
Analogy subspaces across layers and declared nuisance vectors.
Compatibility
Subspace definitions and layer/extraction choices are model-specific.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-06 · Grassmannian velocity · instrument:tool-06

Inspect representationsResearch prototypeCHRONOS empirical lensHow does a reading compare with an empirical reference frame?Source not publicly releasedRead instrument details ↗
Method
Cosine-neighbor comparisons with reference-to-reference baselines.
Output
Reference comparisons under a declared atlas and encoder.
Limit
Requires specific atlas/model assets; its baseline is distinct from a calibrated covariance-Gaussian null.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Identified encoder and empirical reference atlas with its construction metadata.
Inputs
Query representation and reference-to-reference comparisons.
Compatibility
Reference comparisons require the same embedding/preprocessing contract.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-07 · CHRONOS empirical lens · instrument:tool-07

Inspect representationsResearch prototypeDark-matter censusWhat does a population spectrum contain under a stated construction?Source not publicly releasedRead instrument details ↗
Method
Known-answer synthetic bodies and cached population comparisons.
Output
Norm/spectrum diagnostics and refusal records.
Limit
Population definitions, width sensitivity and cache custody constrain interpretation. Acquisition code is not a passive inspector.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
A population definition, identified cache or explicit synthetic construction.
Inputs
Population matrices, norm/width choices or known-answer synthetic bodies.
Compatibility
A model roster is not a completed study; a natural spectrum needs a bound cache.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-14 · Dark-matter census · instrument:tool-14

Compare and transport

Compare and transportExploratory research sourceTranslation lensWhat symbol/context structure can a local induction method recover?Source not publicly releasedRead instrument details ↗
Method
Co-occurrence and PMI-based grammar induction.
Output
Symbol/context and grammar outputs.
Limit
Do not treat this source as a binding to a different model-training study.
Maturity
Exploratory research source
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Paired concept inputs, compatible encoders/taps and fit/test separation.
Inputs
Paired representations and the instrument/readout to transport.
Compatibility
Dimensions, rank, extraction recipe and coordinate conventions must be recorded.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-09 · Translation lens · instrument:tool-09

Compare and transportResearch prototypePortable probesCan a declared linear readout be carried into an evaluation pipeline?Source not publicly releasedRead instrument details ↗
Method
Validate input schemas and score pooled representations with a linear probe.
Output
Scores and input-admission diagnostics.
Limit
A valid artifact schema does not establish semantic accuracy or calibrated self-report.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Receiver residual access and a version-bound ingress/probe package.
Inputs
A bound probe, selected residual tap and receiver configuration.
Compatibility
Static readout ingress does not establish reliable causal intervention.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-10 · Portable probes · instrument:tool-10

Compare and transportDocumented pilot and research sourceAtlasHow well does a fitted transport map carry static probes?Source not publicly releasedRead instrument details ↗
Method
Fit/stitch transport with held-out evaluation and matched orientation controls.
Output
Stored transport decisions and comparison receipts.
Limit
Bounded historical readouts; original activations/acquisition are incomplete. This differs from the public basin-atlas visualization.
Maturity
Documented pilot and research source
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Source/target feature matrices, atlas/stitch recipe and independent evaluation membership.
Inputs
Paired fit representations, probes and held-out evaluation records.
Compatibility
Recorded static transfer and its orientation floors are method-specific.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-11 · Atlas · instrument:tool-11

Compare and transportResearch prototypeWhitening portDoes covariance conditioning change a transport comparison?Source not publicly releasedRead instrument details ↗
Method
Shrinkage whitening, alignment, recoloring and probe evaluation.
Output
Transport statistics and matched-rotation controls.
Limit
Fit/holdout separation is a caller responsibility. These controls are not full covariance-Gaussian surrogates.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Fit covariance, shrinkage/whitening choices and independent evaluation inputs.
Inputs
Source/target covariance and representations with a readout to port.
Compatibility
Random-rotation floors are distinct from full covariance-Gaussian surrogates.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-12 · Whitening port · instrument:tool-12

Compare and transportResearch prototypeShared dimensionHow much held-out operator structure is shared?Source not publicly releasedRead instrument details ↗
Method
Fit-only whitening, ridge operators, correspondence shuffles and width checks.
Output
Explained-operator statistics and plateau/refusal records.
Limit
A correspondence shuffle answers a different question from a Gaussian surrogate. Insufficient draws or absent plateaus should be reported.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Recorded fit/held-out blocks, width sequence, null draws and plateau rule.
Inputs
Paired operators/representations and declared width/control settings.
Compatibility
Insufficient draws or absent plateaus require refusal; controls must match the statistic.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-13 · Shared dimension · instrument:tool-13

Test causal interpretations

Test causal interpretationsResearch prototypeCartography Gaussian dual nullDoes a geometric statistic survive a covariance-aware comparison?Source not publicly releasedRead instrument details ↗
Method
Construct covariance-matched surrogates and separate shuffle/order controls.
Output
Statistic and null comparisons.
Limit
The recovered implementation has version-specific limits; revised paper statistics are not all bound to this source.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Population matrix, SVD/covariance recipe, matched controls and statistic definition.
Inputs
Recorded population representations and specified Gaussian/control constructions.
Compatibility
Recovered implementation and later paper statistics have distinct version bindings.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-04 · Cartography Gaussian dual null · instrument:tool-04

Test causal interpretationsResearch source; retired formulation retainedCausal-excess controlsWhich invariants should a causal comparison preserve?Source not publicly releasedRead instrument details ↗
Method
Construct controls that retain specified geometric properties.
Output
Control vectors and declared invariants.
Limit
Control construction alone is not evidence of causal effect. The CT0-A formulation was retired without a model result.
Maturity
Research source; retired formulation retained
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
An explicit invariant contract and evidence binding the actual scored vectors.
Inputs
Vectors/operators and the geometric invariants each control must preserve.
Compatibility
Retired formulations and source availability do not authorize a causal success claim.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-16 · Causal-excess controls · instrument:tool-16

Calibrate measurements

Calibrate measurementsResearch prototypeCalibration probeCan generated anchors support a useful instrument boundary?Source not publicly releasedRead instrument details ↗
Method
Generator-driven anchors with novelty/separation heuristics and gates.
Output
Anchor diagnostics and gate conditions.
Limit
An active harness, not an offline verifier or an established general calibration result.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Compatible endpoint or internal readout recipe, battery and feature contract.
Inputs
Battery items, responses/features, labels and configuration.
Compatibility
Endpoint output features and internal activation features are different instruments.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-08 · Calibration probe · instrument:tool-08

Calibrate measurementsResearch prototypePort certificateDoes a trained transport statistic exceed its declared floor?Source not publicly releasedRead instrument details ↗
Method
Record trained-minus-floor contrasts, uncertainty and refusal conditions.
Output
Decision and uncertainty receipts.
Limit
Stored summary uncertainty is not equivalent to a paired resampling analysis of original predictions.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Trained and matched-floor results, evaluation membership and uncertainty method.
Inputs
Versioned readout scores/floors and available prediction or summary records.
Compatibility
Summary-only AUROC uncertainty differs from prediction-level paired resampling.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-15 · Port certificate · instrument:tool-15

Calibrate measurementsDocumented synthetic calibrationRKC0 summaryDoes a decision rule behave correctly in known-answer worlds?Source not publicly releasedRead instrument details ↗
Method
Synthetic decision-calibration summaries and interval accounting.
Output
Stored decision and uncertainty records.
Limit
Decision arithmetic does not admit the full natural-model pipeline or establish a biological/causal interpretation.
Maturity
Documented synthetic calibration
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Access to the retained synthetic summary, seal and governing decision rule.
Inputs
Existing planted/matched-null summary records and scope seal.
Compatibility
Stored decision arithmetic is not a complete natural-model pipeline.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

Retained summary/commitment description; no public executable release implied

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-17 · RKC0 summary · instrument:tool-17

Calibrate measurementsDesign commitment; completion unboundTC2 source/seed commitmentWhat calibration design and seeds were committed?Source not publicly releasedRead instrument details ↗
Method
Versioned source and synthetic-world seed commitment.
Output
Design/commitment provenance.
Limit
A committed plan is not a completion receipt. No current job-state or finished-result claim is made.
Maturity
Design commitment; completion unbound
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Access to the versioned source/seed commitment and any separately recovered terminal record.
Inputs
Existing source/seed protocol metadata; terminal evidence where available.
Compatibility
A commitment or historical in-flight note is not proof of completion.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

Retained summary/commitment description; no public executable release implied

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-18 · TC2 source/seed commitment · instrument:tool-18

Calibrate measurementsResearch data builderAxis builderWhich labeled inputs and holdouts define the assay?Source not publicly releasedRead instrument details ↗
Method
Build declared axis corpora and evaluation inputs.
Output
Axis rows, input identities and split-related records.
Limit
Recoverable inputs are not recovered model activations or original effects. Dataset redistribution terms require separate review.
Maturity
Research data builder
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Identified dataset revisions, builder dependencies and explicit split membership.
Inputs
Named corpus sources, labels, axis construction and held-out membership.
Compatibility
Dataset terms and weak-label/ordering choices require separate review.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-21 · Axis builder · instrument:tool-21

Study feedback

Study feedbackResearch prototypeCalibration dataWhat does a feedback/self-report calibration assay actually learn?Source not publicly releasedRead instrument details ↗
Method
Calibration data construction, LoRA training and controlled evaluation.
Output
Calibration and control summaries.
Limit
Natural-model self-report calibration is not established; transcript shortcuts and failed or marginal results remain relevant.
Maturity
Research prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Model/checkpoint access, calibration data, feedback feature contract and matched control arms.
Inputs
Versioned feedback/calibration data, model settings and evaluation controls.
Compatibility
Training loss or a transcript shortcut does not establish truthful natural self-report.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-19 · Calibration data · instrument:tool-19

Audit evidence

Audit evidenceResearch integrity prototypeCrossing spineCan stored decisions and receipt accounting be replayed?Source not publicly releasedRead instrument details ↗
Method
Model-free integrity and bounded receipt replay.
Output
Decision, schema and integrity checks.
Limit
Integrity checks do not certify truth, natural capability or physical containment.
Maturity
Research integrity prototype
Access
Source not publicly released
License
No public release license established in this catalog.
Prerequisites
Versioned schemas, payloads, accounting rules and verification keys where required.
Inputs
Existing receipts, ledger/accounting records and source contracts.
Compatibility
Integrity/replay checks do not attest truth, capability or physical containment.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for source inspection or stored-record review. Model-access methods may require substantial compute; endpoint/provider charges are configuration-specific. No cost estimate or free execution is promised.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Read the method →Link to this instrument →

tool-20 · Crossing spine · instrument:tool-20

Audit evidencePublic utility; private review wrappernulltest.py and the local fixed-battery reviewerCan benchmark labels be recovered from the input surface?Public upstream source; wrapper not releasedRead instrument details ↗
Method
Out-of-fold surface probes with a permutation-based decision rule.
Output
Surface-floor verdict, effect and test diagnostics.
Limit
A PASS is limited to the probes, sample size and test configuration. The local review wrapper is not itself a published release.
Maturity
Public utility; private review wrapper
Access
Public upstream source; wrapper not released
License
Apache-2.0 for the public honesttools source only; local wrapper release is separate.
Prerequisites
Public nulltest source, Python standard library and an authorized labeled JSON battery.
Inputs
A JSON battery with text, binary labels and optional strata; no model access needed.
Compatibility
Specify text/label/stratum fields; power and family/permutation settings constrain interpretation.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for the specific utility, with runtime dependent on input size and test settings. Other charter or Photon service actions have separate access/cost contracts.

Usage example and scope

Public README source-checked at the cited revision; not executed in this site review

python3 nulltest.py battery.json --json

Use only after permission to compute on the chosen battery. Standard-library local computation; no model/API call. A PASS has stated power and feature limits.

Example source or method →
Method and evidence →Inspect public source →Link to this instrument →

tool-22 · `nulltest.py` and the local fixed-battery reviewer · instrument:tool-22

Audit evidencePublic source utilityclaim_gate.pyAre published numbers attached to declared sources?Public sourceRead instrument details ↗
Method
Lint numerical annotations against a registry and strictness policy.
Output
Unannotated-number and source/battery warnings or failures.
Limit
A provenance linter does not validate the science. Warning/strict modes and skipped checks affect its coverage.
Maturity
Public source utility
Access
Public source
License
Apache-2.0 for the public honesttools source only; local wrapper release is separate.
Prerequisites
The public claim-gate source, registry/battery records and its expected adjacent page tree.
Inputs
Page numerical annotations and versioned registry/floor records.
Compatibility
A bare checkout lacks the site tree expected by the source; warnings and strict failure differ.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for the specific utility, with runtime dependent on input size and test settings. Other charter or Photon service actions have separate access/cost contracts.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Inspect public source →Link to this instrument →

tool-23 · `claim_gate.py` · instrument:tool-23

Audit evidencePublic source utilitiesrender_audit.py and nulltest_parity.shDo a rendered result and selected implementations agree under their check?Public sourceRead instrument details ↗
Method
Stale-render checks and bounded Python/JavaScript fixture parity.
Output
Render/fixture comparisons.
Limit
Fixture parity does not establish universal implementation equivalence. Some modes write outputs.
Maturity
Public source utilities
Access
Public source
License
Apache-2.0 for the public honesttools source only; local wrapper release is separate.
Prerequisites
Public render/parity source, expected adjacent site tree and compatible Python/JavaScript runtimes.
Inputs
Audit result/manifest records or declared parity fixtures.
Compatibility
Parity checks bounded fixtures and precision; rendering may write unless check mode is used.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for the specific utility, with runtime dependent on input size and test settings. Other charter or Photon service actions have separate access/cost contracts.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Inspect public source →Link to this instrument →

tool-24 · `render_audit.py` and `nulltest_parity.sh` · instrument:tool-24

Audit evidencePublic source contract; experimental componentshonest.py / contract.pyHow does a packaged tool declare status, tests and refusal?Public sourceRead instrument details ↗
Method
A command registry and tool contract for stable/experimental use.
Output
Tool status and contract outputs.
Limit
Release-specific gate behavior must be checked; packaging does not certify every experimental implementation.
Maturity
Public source contract; experimental components
Access
Public source
License
Apache-2.0 for the public honesttools source only; local wrapper release is separate.
Prerequisites
Public command/contract source and the specific tool release status.
Inputs
Tool identity, contract/FPR document and command-specific input.
Compatibility
Stable/experimental states and known green-gate limitations must be read per tool.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for the specific utility, with runtime dependent on input size and test settings. Other charter or Photon service actions have separate access/cost contracts.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Inspect public source →Link to this instrument →

tool-25 · `honest.py` / `contract.py` · instrument:tool-25

Audit evidencePublic verification utility and proposal ledgerSigned verification / broader charterWho signed these bytes, and what does the tool contract cover?Public source; additional tools varyRead instrument details ↗
Method
Signature verification plus a public ledger of available and proposed tools.
Output
Authenticity checks and release/proposal status.
Limit
Signatures do not establish truth. Transparency witnessing/co-signing and proposed tools are separate maturity questions.
Maturity
Public verification utility and proposal ledger
Access
Public source; additional tools vary
License
Use the license in each public source artifact; the public honesttools repository is Apache-2.0. Other artifacts are not covered by that license.
Prerequisites
Versioned verifier/specification, signed payload and the issuer public key.
Inputs
Exact signed bytes, signature/public key and governing verification spec.
Compatibility
Different proposed or partial tools have separate availability; no blanket shipped status.
Replication boundary
See the evidence record; executable source, packaging checks and retained summaries are not independent scientific replication.
Reproducibility
Record the exact artifact version, input identity, extraction/settings, split/control definitions, seed and output receipt. Distinguish source inspection, stored-summary replay and independent experiment replication.
Authorization boundary
Use requires separate authorization for the selected inputs, compute and actions. This catalog does not authorize experiments, model/provider calls, training, publication or private-data transfer.
Costs
Local CPU/time for the specific utility, with runtime dependent on input size and test settings. Other charter or Photon service actions have separate access/cost contracts.

Usage example and scope

No public runnable command verified for this entry

Example source or method →
Method and evidence →Inspect public source →Public verifier →Link to this instrument →

tool-26 · Signed verification / broader charter · instrument:tool-26

26 instruments · complete public catalog

These entries enumerate the existing shelf; they are not independent tools deployed as services or independently validated results. Read the machine-readable catalog, researcher agent guide or public utility documentation.