Research registry
Claim ledger
Claim dispositions and falsifiers generated from the corpus authority.
Priority records
2v1 · 2026-07-22
Tarski Coherence
Which degree-1 Hodge-Tarski statements hold for finite lattices beyond vector spaces?
- Formally verified results
The degree-1 Hodge-Tarski conjecture is false as stated and admits an exact graph-theoretic characterization in the paper's model.
A countermodel to a manuscript theorem under its stated hypotheses, or a mismatch between a headline statement and the clean Lean build.
v2-draft · 2026-07-22
BABEL v2
Do structural signals predict execution failure on independently authored workflows?
- Conjectural
No independent workflow evidence currently validates structural signals as predictors of execution failure.
Failure to improve held-out AUROC by 0.05 with a positive bootstrap interval, or failure of the partner gates, ends the predictive claim.
Supporting records
13draft-2026-07 · 2026-07-22
Compositional Accountability
When do local accountability records compose into one challenge-time account across owners?
- Empirically demonstrated
The source-only finite profile computes an exact model-relative query-answerability margin and replayable erasure certificate; it does not assess occurrence truth, non-equivocation, authority, remedy, or external validity.
A differential mismatch, an accepted tampered certificate, a case where control-domain or occurrence binding is ignored, or any public claim promoting the model-relative result to occurrence truth invalidates the profile.
v1 · 2026-07-22
Column-Matroid Backbone
Which matroid identity underlies the disclosure deficit?
- Not assessed
The scoped fee is a column-matroid corank under the paper's construction and hypotheses.
A counterexample under the exact construction and stated hypotheses.
v1 · 2026-07-22
Signed-Incidence Structure
Which field-independent rank results follow from signed-incidence structure?
- Formally verified results
The completed archived Lean source proves signed_incidence_det_in_unit; field independence is the written consequence through nonzero minors.
A counterexample under the paper's exact hypotheses.
v2 · 2026-07-22
A Witness Logic for Semantic Composition
Which numerical invariant follows from the seam-independent disclosure axioms?
- Formally verified results
Numerical uniqueness over the abstract DoctrineCarrier is machine-checked.
A model satisfying the abstract axioms but violating the checked theorem, or failure of the claimed concrete identification.
v1 · 2026-07-22
SCPI
How does predicate invention decompose into topological, model-theoretic, and definability gates?
- Formally verified results
Finite Z/2 examples and finite-indexed set-valued gluing are machine-checked under the exact declarations.
- Falsified
conservativity_descent supplies a general semantic descent theorem beyond its h_local assumption.
- Falsified
Beth.lean machine-verifies semantic Beth definability for sites or extracts an interpolant.
- Not assessed
The prose geometric or general Beth-for-sites assembly is established.
A counterexample to the retained formal core under its exact abstractions, or any public surface using Beth.lean as semantic theorem or extractor evidence.
v1 · 2026-07-22
Bridge
Can bilateral validity coexist with failure to specify a coherent bridge?
- Empirically demonstrated
On the disclosed hidden-repair study, 14 of 15 classifications matched the predicted repair type exactly, one matched partially, and none missed.
- Not assessed
The study independently establishes protected-language non-conservativity, non-definability, or model policy canonization.
Reproduction failure under the disclosed hidden-repair protocol falsifies the retained baseline; the untested Gate 2, Gate 3, and canonization claims receive no positive status.
v1 · 2026-07-22
Seam
Which protocol records local interface claims that do not compose automatically?
- Empirically demonstrated
The protocol and tooling are implemented; this does not establish universal safety or recourse.
Failure of the stated implementation tests or an unsupported universal consequence claim.
v8 · 2026-07-22
Bitcoin After Money
Which settlement requirements follow from autonomous coordination without a trusted issuer?
- Conjectural
Constitutional argument, not an empirical or formally verified paper.
A demonstrated substitute satisfying the paper's stated settlement requirements.
v1 · 2026-07-22
Coherence Cliff
How do the structural measures scale in the frozen experimental regime?
- Empirically demonstrated
Scaling result is limited to the frozen experimental regime.
Reproduction failure on the frozen protocol or a cheap baseline matching the structural method.
v1 · 2026-07-22
Local Interface Descriptions Do Not Compose
Does representation affect access to non-local interface predicates?
- Empirically demonstrated
Scoped constructed demonstration, not evidence of population prevalence.
Failure to reproduce the constructed instances.
v1 · 2026-07-22
The Stitching Defect
Which defect appears when two locally analyzed systems are stitched across a boundary?
- Empirically demonstrated
Scoped result retained as a technical report.
Reproduction failure under the stated protocol.
v1 · 2026-07-22
The Composition Tax
What does the live proxy report about the cost of semantic composition?
- Empirically demonstrated
Applied results are limited to the disclosed proxy, schemas, and trials.
Reproduction failure or a frozen cheap baseline matching the reported result.
v1 · 2026-07-22
Local Validity Does Not Compose
Can locally valid interfaces fail to support a globally coherent account?
- Empirically demonstrated
Scoped constructed and empirical result only.
Reproduction failure under the stated protocol.
Frontier records
6draft-2026-07 · 2026-07-22
The Disclosure Deficit
Which fee-family results remain after removing the topology, safety, and prediction claims?
- Formally verified results
Specified rank identities have Lean support; field independence uses the checked total-unimodularity step and a written nonzero-minor argument.
- Not assessed
The minimum-disclosure-set product count has a written correction; no matching machine-checked declaration is located.
- Falsified
Repair is operationally sparse under integer costs in the stated box.
- Falsified
The deficit predicts execution outcomes, or carries a safety or topological reading.
Any included result that depends on cycle necessity, topology branding, execution prediction, or a false repair-basis statement must be removed.
v1-draft · 2026-07-22
SHEAF
Does the local-to-global sheaf framework extend to heterogeneous agents?
- Conjectural
Heterogeneous-agent extension under development; the reset does not depend on it.
Counterexamples to the stated conjectures or failure to separate the model from adjacent prior art.
v1 · 2026-07-22
Compositional Incoherence Is a Parity
When is compositional incoherence invisible to bounded-radius certification?
- Conjectural
Hardness result remains paper-level pending independent review.
A flaw in the reduction or an algorithm violating the stated lower bound.
v0.1-methods-draft · 2026-07-22
The Golden Gate
Can semantic-control systems be evaluated for accepting, refusing, requesting evidence, preserving choice, and abstaining for the correct checkable reason?
- Empirically demonstrated
The Golden Gate methods and captive qualification machinery are implemented and reproducible from the disclosed packet.
- Not assessed
The system has passed reviewer-originated external evaluation or independent implementation parity.
A custody path that permits author access before submission, a zero-import verifier that depends on Bulla, an unsafe accepted Golden case, or any external-result claim without reviewer-controlled evidence invalidates the method claim.
v0.1-internal-gate · 2026-07-22
Interpolant Envelope
When can a finite disputed predicate be safely defined in shared vocabulary, partially certified, or disproved by same-reduct countermodels?
- Formally verified results
In the declared finite FRSL-1 semantics, full definitions, partial RELY and REFUSE regions, and same-reduct non-definability certificates can be independently checked.
An emitted package that passes the standalone checker while violating its finite semantic obligations, a non-definability certificate whose paired expansions do not share the protected reduct, or a mismatch between the Lean statement and executable abstraction blocks the claim.
v0.4-internal-draft · 2026-07-22
No Free Precedent
Under what explicit reason, authority, scope, closure, and epoch may one forum finding govern a later case?
- Formally verified results
Case labels on a proper subset do not determine an unseen case absent an explicit hypothesis restriction.
- Empirically demonstrated
Authorized precedent compounded on the frozen team-authored planted-rule lineages under bounded-exact closure.
- Not assessed
Precedent compounding generalizes to foreign meanings or creates economic value.
A case-only finding that verifies as a general precedent, an application without separate adoption and applicability, same-epoch refusal erasure, stale precedent reuse, or ambient authority invalidates the typed-flow claim.
Historical records
9v0.1.0 · 2026-07-22
BABEL v1
Does the frozen benchmark support reproducible compositional semantic-failure tests?
- Empirically demonstrated
The frozen benchmark and its disclosed evaluations remain reproducible within their stated protocol.
- Falsified
Internal structural labels establish prediction of real execution failure.
Reproduction failure under the frozen v0.1 protocol falsifies the retained benchmark claim; no execution-prediction claim survives.
v5 · 2026-07-22
The Coherence Fee gateway essay
Does a local-to-global disclosure deficit predict adverse execution outcomes?
- Falsified
Runtime-failure prediction from fee magnitude.
- Formally verified results
The audited component-count and omission identities hold under their stated hypotheses.
Not applicable to this historical record.
v1 · 2026-07-22
The Coherence Fee Paper I
Does a hierarchical rank deficit represent hidden convention dimensions and prescribe repair?
- Falsified
Historical source only.
Not applicable to this historical record.
v1 · 2026-07-22
The Witness Gram Paper III
Do spectral invariants determine substitutable disclosure choices and repair bases?
- Falsified
Historical source only; valid rank identities may be reused after line-by-line audit.
Not applicable to this historical record.
v1 · 2026-07-22
Witness Geometry Beyond Scalar Fee
Does repair geometry yield sparse operational prescriptions beyond a scalar deficit?
- Falsified
Historical source only.
Not applicable to this historical record.
v1 · 2026-07-22
Coherence Types
Is the proposed coherence calculus sound under its refinement-type discipline?
- Falsified
Historical source only.
Not applicable to this historical record.
v1 · 2026-07-22
The Linear Communication Bottleneck Conjecture
Does the spectral-gap trichotomy imply a linear communication bottleneck?
- Falsified
Surviving scoped result: random-encoder Procrustes edges have exact Haar marginals and exact adjacent-edge pairwise independence.
A counterexample to conditional Haar-ness or adjacent-edge independence falsifies the surviving lemmas; no spectral claim survives.
v1 · 2026-07-22
The Physical Coherence Fee
Does the disclosure deficit extend to a physical resource theory?
- Conjectural
Archived during the 90-day reset; no new research family is authorized.
Not applicable to this historical record.
v1 · 2026-07-22
Interpretability Frontier
Does component-level interpretability recover composition-level semantics?
- Empirically demonstrated
In the paper's controlled cyclic-composition regime, edge-local interpretability features did not recover the global structural signal.
- Not assessed
The reported result establishes a general boundary for all interpretability systems.
Not applicable to this historical record.