Vocabulary

Every status term used anywhere on this site resolves to a definition here. A term with no definition cannot be rendered — the build fails instead.

Ontology version: 1.3.0
Issued: 2026-09-11
Machine-readable: /api/ontology.json

Every status term used anywhere on this site MUST resolve to a definition in this document. Terms are versioned. Definitions are never silently edited; a changed meaning requires a new ontology version and a recorded reason. Version 1.1.0 adds availability, release_state and access as separate dimensions. They describe software and resources, never evidence: a thing can be fully available and scientifically unqualified, or qualified and unavailable, and collapsing the two is the most common way a research site misleads. Version 1.2.0 adds development_state: where a thing sits in research and construction, which is a different question from whether it is available, whether it is scientifically qualified, and what access it needs. A thing in discovery is not a product awaiting release, and it is not a failed product. Version 1.3.0 adds the future machine-operation vocabulary: an authority ladder, the states an operation passes through, and typed stop reasons. None of it is active. No delegation exists, so every rung is currently held by nobody.

Qualification

What standing does the laboratory itself assert for this object?

FROZEN_ADOPTED: Frozen / adopted — Empirically derived, subjected to falsification attempts, frozen against further edit, and adopted as an operating standard inside the laboratory.
Does not imply: external validation; independent replication
QUALIFIED: Qualified — Passed a documented qualification procedure whose criteria were fixed before execution, across the full domain the claim asserts.
Requires: evidence_level>=CONTROLLED_REPEATED; provenance_completeness!=INCOMPLETE; at least one executed falsifier
QUALIFIED_AT_CURRENT_SCOPE: Qualified at current scope — Passed qualification only inside an explicitly enumerated tested scope. Behaviour outside that scope is not asserted.
Requires: non-empty tested scope; non-empty excluded or untested scope
EXPERIMENTAL: Experimental — Observed under conditions the laboratory controls, but not qualified. May be wrong.
FRONTIER_BOUNDED: Frontier-bounded — The limiting factor is the current measurement or apparatus frontier, not the phenomenon. The bound is a property of our instruments.
UNOBSERVED: Unobserved — Stated as a structured object so it can be tested and cited, but never observed. Carries no evidential weight.
NOT_PRODUCTION_QUALIFIED: Not production qualified — May be scientifically informative and is explicitly unfit for operational or production reliance.

Lifecycle state

Is this the laboratory's current position, or history?

CURRENT: Current — The laboratory's present position for this claim identity.
SUPERSEDED: Superseded — Replaced by a later version. The record remains permanently readable and citable. Supersession is not deletion and not retraction.
WITHDRAWN: Withdrawn — Removed from the laboratory's asserted positions for a stated reason. It may never return to CURRENT; a later claim must be a new version with new evidence.
FALSIFIED: Falsified — A preregistered falsifier executed and the claim failed it. This is a scientific result, not an error to be hidden.
HISTORICAL: Historical — Retained for the record only; the laboratory asserts nothing about it today.

Evidence level

What kind of observation stands behind it?

ASSERTED_ONLY: Asserted only — No observation recorded. Structure without evidence.
SINGLE_OBSERVATION: Single observation — One run, no control, no repetition.
CONTROLLED_SINGLE: Controlled, single — One execution with a stated control or negative control.
CONTROLLED_REPEATED: Controlled, repeated — Repeated executions with controls and a stated stopping rule.
PREREGISTERED_CONTROLLED: Preregistered, controlled — Hypothesis, endpoints, exclusions and analysis were timestamped and digested before execution.

Causal level

How strong is the causal language allowed to be?

NONE_ASSERTED: No causal assertion — Descriptive only.
ASSOCIATIONAL: Associational — A relationship was observed. No intervention was performed. Causal wording is not authorised.
INTERVENTIONAL: Interventional — A variable was deliberately manipulated with a control condition.
Requires: an experiment object with a stated control
MECHANISTIC: Mechanistic — Intervention plus an articulated and separately tested mechanism.

Replication state

Did anyone other than the original run get the same result?

NONE: None — Not reproduced and not replicated.
INTERNAL_REPRODUCTION: Internal reproduction — The same actors re-ran the same implementation on the same data and obtained the agreed result. This is NOT independent replication; it tests our determinism, not the finding.
INDEPENDENT_REPLICATION_PARTIAL: Independent replication (partial) — An independent party replicated a proper subset of the claim components, or matched only some equivalence criteria. Displaying this as full replication is a defect.
INDEPENDENT_REPLICATION_FULL: Independent replication (full) — An independent party, independent implementation and independent data satisfied every predefined equivalence criterion for every claim component.

Measurement validity

Does the instrument measure what the claim is about?

UNVALIDATED: Unvalidated — The measurement instrument has not been shown to measure the construct.
CONSTRUCT_UNCERTAIN: Construct uncertain — The instrument is reliable, but its relationship to the construct the claim names is disputed or unestablished.
VALIDATED_IN_SCOPE: Validated in scope — Instrument validity was demonstrated inside the tested scope only.

Contamination / leakage

Could the result be an artefact of the data or the apparatus?

UNASSESSED: Unassessed — No contamination or leakage analysis was performed. Absence of a finding here is not evidence of cleanliness.
SUSPECTED: Suspected — A specific contamination or leakage pathway is suspected and unresolved.
KNOWN_PRESENT: Known present — A contamination or leakage pathway is confirmed and its effect on the claim is stated.
ASSESSED_NONE_KNOWN: Assessed, none known — A documented analysis ran and found no contamination pathway. Bounded by what the analysis could see.

Provenance completeness

Can the path from raw work to this public record be traced?

INCOMPLETE: Incomplete — At least one step between the original work and this record is unrecorded.
PARTIAL: Partial — The chain is recorded but at least one step is a sanitised attestation rather than an inspectable input.
COMPLETE: Complete — Every transformation from input to published record is recorded with input and output digests.

Component support

For each separable part of the claim, what actually backs it?

DIRECT: Directly supported — An evidence object measures this component itself.
INDIRECT: Indirectly supported — Supported by inference from evidence about something else.
SCOPE_LIMITED: Supported, scope-limited — Directly supported inside a narrower domain than the component states.
UNREPLICATED: Supported, unreplicated — Directly supported by a single unreplicated source.
CONTRADICTED: Contradicted — At least one evidence object points the other way, and the conflict is unresolved.
UNMEASURED: Unmeasured — No instrument in the laboratory can currently measure this component.
ASSUMPTION_DEPENDENT: Assumption-dependent — Holds only if a stated, untested assumption holds.
UNSUPPORTED: Unsupported — Stated in the claim and backed by nothing. Present so it is visible, not because it is believed.

Evidence debt type

What does the laboratory still owe this claim?

MISSING_REPLICATION: Missing independent replication — No independent party has attempted the claim.
MISSING_OOD_CHALLENGE: Missing out-of-distribution challenge — Never tested outside the distribution it was developed on.
MEASUREMENT_GAP: Measurement gap — A named quantity in the claim has no valid instrument.
CONTAMINATION_UNRESOLVED: Unresolved contamination — A suspected leakage pathway has not been ruled out.
UNTESTED_BOUNDARY: Untested boundary — A scope edge asserted in the claim was never probed.
INCOMPLETE_PROVENANCE: Incomplete provenance — A step in the chain is unrecorded.
UNEXECUTED_FALSIFIER: Unexecuted falsifier — A registered falsification test has never been run.
DATA_UNAVAILABLE: Data unavailable — Evidence exists privately and cannot lawfully be released.
UNRESOLVED_CONTRADICTION: Unresolved contradiction — Two evidence objects disagree and neither has been retired.
RELEASE_RESTRICTED: Release restricted — Licensing, security or provider terms block publication of the supporting evidence.

Release classification

What may lawfully be published?

PUBLIC: Public — Cleared for unrestricted publication.
PUBLIC_SANITIZED: Public (sanitised) — Published only as a derived record whose private inputs were removed under a recorded transformation.
DEMO_FIXTURE: Demo fixture — Not science. An exercise record used to test this infrastructure. Excluded from every real scientific index, feed and API.
EMBARGOED: Embargoed — Cleared in substance, blocked until a stated time or event.
INTERNAL: Internal — Never published.
PROTECTED: Protected — Isolated confirmation material. Never published, never referenced in a way that discloses content.
SECURITY_SENSITIVE: Security sensitive — Publication would create operational risk.
LICENSE_RESTRICTED: Licence restricted — Third-party terms forbid redistribution.
PRIVATE_INFRASTRUCTURE: Private infrastructure — Operational configuration. Never published.

Reproduction kind

What was actually re-done?

ARTIFACT_VERIFICATION: Artifact verification — Bytes match a published digest. Says nothing about whether the science is right.
COMPUTATIONAL_REPLAY: Computational replay — The same code on the same inputs produced the same outputs.
EXPERIMENTAL_REPRODUCTION: Experimental reproduction — The same protocol re-executed by the same laboratory.
INDEPENDENT_REPLICATION: Independent replication — A different party, implementation and data satisfied predefined equivalence criteria.

Equivalence criterion

What was agreed, in advance, to count as the same result?

BYTE_EQUALITY: Byte equality — Outputs are identical byte for byte.
TOLERANCE: Numeric tolerance — Outputs agree within a stated tolerance.
DISTRIBUTIONAL: Distributional equivalence — Output distributions agree under a stated test and threshold.
INVARIANT_PRESERVATION: Invariant preservation — A named invariant holds in both runs.
STATE_TRANSITION: State transition — The system reached the same named state.

Availability

Can a member of the public actually get this today?

AVAILABLE: Available — Publicly obtainable and usable right now, by anyone, without an account.
LIMITED: Limited — Partly obtainable. Some component, source or platform is not yet public, and the page says which.
PREVIEW: Preview — Reachable but not stable. Expect it to change without a deprecation window.
PLANNED: Planned — Intended, with the shape of the thing decided. Nothing exists to obtain.
NOT_YET_AVAILABLE: Not yet available — Named because it is real inside the programme, and nothing about it has been cleared for public release. No description, no artifact, no date.
RETIRED: Retired — Was available; no longer offered. Prior releases remain addressable where they were published.

Release state

Is there a release, and is it the one to use?

UNRELEASED: Unreleased — Nothing has been released. This is the state of anything that has not yet produced a public artifact, and it is not a statement that a release was cancelled.
PREVIEW: Preview — A release exists and is not stable. Expect it to change without a deprecation window.
CURRENT: Current — The release the laboratory points at today.
SUPERSEDED: Superseded — A later release exists. This one remains downloadable and addressable.
DEPRECATED: Deprecated — Still working, scheduled to stop. The replacement is named.
WITHDRAWN: Withdrawn — Pulled for a stated reason. Remains addressable so anyone who used it can find out why.

Access

What is required to reach it?

PUBLIC: Public — No credential. No account. No request.
AUTHENTICATED: Authenticated — A credential would be required. No such credential exists today.
RESTRICTED: Restricted — Release is constrained by licence, security or provider terms.
INTERNAL: Internal — Never published.

Development state

Where is this in research and construction?

DISCOVERY: Discovery — Under active research. What the thing is, what it should do, and whether it takes this form at all are open questions. No architecture, interface or capability follows from a name at this stage.
DESIGNING: Designing — A form has been chosen and is being worked out. Still subject to change from evidence.
EXPERIMENTAL_INTERNAL: Experimental (internal) — Exists and runs inside the programme. Not a product, not public, and not a commitment to become either.
REALIZATION_PENDING: Realization pending — Design settled; construction of a usable form has not been completed.
QUALIFICATION_PENDING: Qualification pending — Built enough to be tested; has not passed the qualification it would need to carry a scientific status.
PUBLICATION_REVIEW_PENDING: Publication review pending — Qualified in substance; awaiting the export, sanitisation, classification, licence and release-policy controls that govern anything public.
PUBLICATION_CLEARED: Publication cleared — Cleared for public release. Publication may not have happened yet.
AVAILABLE: Available — Public and obtainable. Development, for this version, is complete.
LIMITED: Limited — Partly public. Some component remains unreleased, and the record says which.
NOT_YET_AVAILABLE: Not yet available — Nothing about it is public. Used only where a more specific development state is not itself publishable.
SUPERSEDED: Superseded — A later version replaced it. It remains addressable.
WITHDRAWN: Withdrawn — Published, then pulled for a stated reason. Reachable only by something that was actually released.

Delegated authority

What is an actor permitted to do — which is a different question from whether what they produce is scientifically sound?

OBSERVE: Observe — Read public state. Confers nothing else, and is the only rung that can be held safely by an actor nobody has evaluated.
PREPARE_DRAFT: Prepare draft — Produce a draft record in a private workspace. A draft has no public existence and never appears in an index, feed, API or current pointer.
PROPOSE_UPDATE: Propose update — Submit a draft as a proposal against a named current version. A proposal is an offer, not a change.
ATTACH_APPROVED_EVIDENCE: Attach approved evidence — Link evidence that has already cleared publication review. It does not confer the ability to approve evidence.
VALIDATE: Validate — Run the policy and schema gates against a proposal and receive the result. Validation is a measurement, not a permission: passing it authorises nothing.
REQUEST_PUBLICATION: Request publication — Place a validated proposal in front of whoever holds publication authority. Requesting is not receiving.
EXECUTE_REVERSIBLE: Execute reversible — Perform an action whose complete undo is defined in advance and demonstrated to work. If the rollback path is unavailable, the action is not reversible and this rung does not cover it.
PUBLISH_APPROVED: Publish approved — Make public a package that already carries an explicit publication decision from an authorised human. It confers no ability to make that decision.
SUPERSEDE_APPROVED: Supersede approved — Replace a current version with an approved successor. The prior version remains published and addressable.
WITHDRAW_APPROVED: Withdraw approved — Retract an approved record with a recorded reason. Withdrawal deletes nothing.
RELEASE_APPROVED_SOFTWARE: Release approved software — Publish a software artifact that has cleared security, licence and provenance review, with its digest.

Operation state

Where is a proposed change, and can the public see it?

DRAFT: Draft — Being written. Private to its author. Not addressable publicly.
PROPOSAL: Proposal — Submitted against a named current version. Still not a change.
SHADOW_RESULT: Shadow result — A full rendering of what the change would produce, generated without affecting anything. Retained for comparison and never served publicly.
APPROVED_FOR_PUBLICATION: Approved for publication — An authorised human has decided it may be published. Not yet public.
PUBLISHED: Published — Public, immutable, digested and citable.
SUPERSEDED: Superseded — Replaced by a later version and still served at its original address.
WITHDRAWN: Withdrawn — Retracted with a recorded reason, still served so the record of the retraction survives.
CONFLICT: Conflict — The world moved while the proposal was being prepared. Parked, both states preserved, never auto-merged.
REJECTED: Rejected — Refused by a gate, with the reason recorded as structured data.

Stop reason

Why did a machine operation halt instead of proceeding?

AMBIGUOUS_CLASSIFICATION: Ambiguous classification — Release classification is absent, unrecognised or contradictory. Fails closed to internal.
INSUFFICIENT_AUTHORITY: Insufficient authority — The actor does not hold the rung this operation requires.
EXPIRED_AUTHORITY: Expired authority — The delegation has passed its expiry.
REVOKED_AUTHORITY: Revoked authority — The delegation was revoked. Revocation is immediate and is never retroactively softened.
TARGET_OUT_OF_SCOPE: Target out of scope — The target resource or environment lies outside the delegation's scope.
PROVENANCE_BREAK: Provenance break — The chain from input to output is incomplete or does not resolve.
SECRET_SCAN_FAILURE: Secret scan failure — Material that looks like a credential or key was detected in the proposal.
PROTECTED_DATA_CONTACT: Protected data contact — The operation touches protected or private material.
LICENSE_UNRESOLVED: License unresolved — Ownership or licence is ambiguous. Ambiguous ownership is non-releasable.
SCHEMA_FAILURE: Schema failure — The record does not satisfy its published schema.
SEMANTIC_OVERCLAIM: Semantic overclaim — Public prose asserts more than the structured record supports.
SCIENTIFIC_STATUS_MISMATCH: Scientific status mismatch — The asserted scientific status is not supported by the attached evidence.
UNSUPPORTED_TRANSITION: Unsupported transition — The requested state change is forbidden by the state machine.
STALE_VERSION: Stale version — The expected current version or digest no longer matches. The world moved.
CONFLICTING_PROPOSAL: Conflicting proposal — Another proposal targets the same object and has not been resolved.
MISSING_REQUIRED_EVIDENCE: Missing required evidence — Policy requires evidence that was not supplied.
ROLLBACK_UNAVAILABLE: Rollback unavailable — No demonstrated undo exists, so the action cannot be treated as reversible.
SECURITY_SENSITIVE: Security sensitive — Publication would create operational risk.
UNKNOWN_CONSEQUENTIAL_FIELD: Unknown consequential field — The envelope carries a field the contract does not model. A field nobody understands may be the one carrying the claim.
POLICY_VERSION_MISMATCH: Policy version mismatch — The actor evaluated against a different policy version than the one in force. Actors do not choose the policy they are judged by.
PUBLICATION_AUTHORIZATION_ABSENT: Publication authorization absent — No explicit publication decision from an authorised human is attached.
IDEMPOTENCY_COLLISION: Idempotency collision — The idempotency key was reused with a different payload.