VámossyResearch

Components

Descriptions below are taken from the toolkit README, architecture index, and accompanying design notes. They do not claim capabilities beyond those sources.

Epistemic / integrity

Goodhart auditor

tools/goodhart_auditor.py

An epistemic linter: flags fields/metrics whose name claims a verified property (reviewed, verified) that nothing actually checks.

Knowledge maturity

tools/knowledge_maturity.py

A deterministic evidence-maturity classifier: rates how much evidentiary work stands behind a claim (not whether it is true), with critical gates that quantity cannot buy past.

Decoupling monitor

tools/decoupling_monitor.py

A Goodhart-in-the-wild monitor: watches a reported proxy against an independent truth signal and alerts when the proxy keeps improving while the truth degrades.

Ground-truth auditor

tools/ground_truth_auditor.py

Checks whether the truth signal a decoupling monitor relies on is actually independent of the proxy, so you know whether a decoupling alarm means anything.

Eval-gaming detector

tools/eval_gaming_detector.py

A defensive eval-gaming / sandbagging detector. Flags when a model-evaluation score is decoupled from true capability. Abstract scores only; no techniques.

White-raven governor

tools/white_raven_governor.py

Governor accompanying the raven taxonomy for rare confirming/disconfirming cases.

Raven taxonomy

tools/raven_taxonomy.py

Taxonomy for confirmation and disconfirmation cases used by the white-raven governor.

Decision

Optimal timing

tools/optimal_timing.py

An optimal-stopping timing layer: from a cost structure and an evidence model, it solves the Bayes-optimal act-or-wait boundary on the posterior.

Option space

tools/option_space.py

Option-set integrity governor: Pareto frontier (no blended score), decoy/completeness flags, non-self-approval. Governs the set before the choice.

Governed decision

tools/governed_decision.py

Composes trust, maturity, timing, safety, and authority into one governed decision. Never emits ACTED. Human authority remains the final slot.

Temporal decision seam

tools/temporal_decision_seam.py

The seam at which temporal claims meet human authorization.

Containment

Containment guard

patterns/containment_guard.py

A fail-closed guard that rejects any agent action which is not human-authorized, reversible, bounded, and logged.

Capable-agent cage

tools/capable_agent_cage.py

A fail-closed boundary over capable-agent proposals. The cage's decision space is small enough to enumerate; stress-test results are published separately.

Consolidation

Scientific knowledge-governance pipeline

soi/soi_pipeline.py

One deterministic function that assigns a claim's epistemic-status order (PROVISIONAL → WORKING_BASIS → MULTI_DOMAIN_TESTED → VALIDATED → CANONICAL_CANDIDATE). It orders claims by status, never by truth.

Meta-governors

Determinism governor

tools/determinism_governor.py

Tries to refute a component's determinism claim (repeat, dict-reorder, order-free, inconsistent-raise) plus a source-smell linter.

Dimensional governor

tools/dimensional_governor.py

Generalizes the refutation engine to a finite, declared set of behavioral dimensions (determinism, purity, idempotence, monotonicity, boundedness, order-invariance).

Mathematics / ontology

Fixed-point governor

tools/fixed_point_governor.py

Detects well-founded versus ungrounded governance towers. Ensures governance chains bottom out at human authority.

Dependency graph

tools/dependency_graph.py

Traces root causes and checks that explanations are well-founded rather than circular.

Temporal governor

tools/temporal_governor.py

Enforces tense-to-verifiability mapping so forecasts are not treated as facts.

Math-to-reality map

tools/math_to_reality.py

Governance-layer mapping between mathematical structure and claims about physical or operational reality.

Fractal recursion

tools/fractal_recursion.py

Notes on fractal recursion as used by the toolkit's structural checks.

Fractal prerequisite

tools/fractal_prerequisite.py

Prerequisite structure for the fractal recursion notes.

Layered sciences

tools/sciences_layers.py

Notes on layered sciences and dependency between domains.

Unverifiable claims

Qualia report governor

tools/qualia_report_governor.py

Handles experience-claims where no third-person ground truth exists. The honest verdict is permanently unverifiable, not a failed check.

Applications

Taxonomy builder

tools/taxonomy_builder.py

Builds and validates taxonomies; used in the log-severity case study.

Postmortem integrity

tools/postmortem_infra.py

Checks whether a postmortem is honest in structure — not whether its conclusions are correct.

Telemetry infrastructure

tools/telemetry_infra.py

Telemetry and alerting infrastructure used with timing and decoupling checks.

Temporal telemetry

tools/temporal_telemetry.py

Temporal telemetry accompanying the temporal governor.

Tokenization taxonomy

tools/tokenization_taxonomy.py

Taxonomy and notes on tokenization as a representational proxy.

Words vs numbers

tools/words_vs_numbers.py

Formalization of the gap between verbal claims and numeric checks.

Freedom infrastructure

tools/freedom_infra.py

Infrastructure note on freedom as used in the toolkit's option and authority model.

EM family

tools/em_field.py

Notes on the electromagnetism / energy-matter family of checks.

Patterns

Governed switch

patterns/governed_switch.py

A governed switch pattern for changing regimes without silent self-authorization.