TA-14 PUBLIC GOVERNANCE SHOWCASE · REGISTERED GOVERNANCE OBJECT

ADMISSIBLE COMPUTATION

TA-14 Admissible Computation Architecture · ACA v1.1

Govern before computational commitment. TA-14 ACA asks what must be true before a proposed computational path has sufficient standing to consume resources and create consequence. It separates a request from the workload proposed to satisfy it, places an explicit admissibility and binding boundary before COMMIT, and preserves outcome as a new state rather than inherited authority.

Open permanent Registry record Run Admissible Search Gateway Test in ACA Pilot Lab Open canonical Zenodo publication

10Bounded claims

4Pre-commit states

7Core chain stages

OPENIndependent examination

TA-14 REGISTRY ADMISSION

REGISTERED — NOT CERTIFIED

This public record establishes a dated, attributable, searchable registration of TA-14 ACA Version 1.1 and its declared public information. Registration does not certify effectiveness, legality, safety, compliance, operational suitability, energy savings, environmental benefit, or fitness for execution.

The computational governance chain

ACA makes the proposed workload itself governable before resources are committed.

REQUEST

COMPUTATIONAL PROPOSAL

ADMISSIBILITY

BINDING

COMMIT

COMPUTATION

OUTCOME

ALLOW

HOLD

DENY

ESCALATE

What ACA governs

The architecture separates capability from standing and standing from commitment.

NECESSITY + PURPOSEAsk what objective the work serves and whether the path remains necessary before resources are committed.

SCOPE + PROPORTIONALITYBind data, models, tools, resource class, limits, and stop conditions to the consequence that justified the request.

EVIDENCE + AUTHORITYRequire sufficient present evidence and relevant authority where authority matters; capability alone does not create standing.

CONTINUITY + REVALIDATIONMaterial changes to identity, evidence, authority, scope, purpose, resource conditions, policy, or outcome can invalidate affected standing.

COMMIT BOUNDARYBefore COMMIT, work is proposed and governed. After COMMIT, only admitted resources may be consumed within the binding.

OUTCOME + RESETCompleted computation becomes evidence for a new state. It does not silently authorize the next computational consequence.

Ten registered claims

These are claimant-declared claims preserved by registration. They are not TA-14 certification findings.

ACA-C1Computational capability alone does not establish sufficient standing to commit a workload.

ACA-C2A proposed computational path can be governed before COMMIT.

ACA-C3Admissibility is consequence-bound rather than a universal trust score.

ACA-C4ALLOW, HOLD, DENY, and ESCALATE are explicit computational governance states.

ACA-C5Permission can be bound to purpose, scope, evidence, authority, continuity, model/tool class, resource envelope, and intended consequence.

ACA-C6Material changes relevant to standing require appropriate revalidation.

ACA-C7Completed computation creates an outcome record and new state rather than inherited authority for subsequent computation.

ACA-C8Computational branches can be governed for necessity, scope, proportionality, continuity, and stop conditions.

ACA-C9Gate cost can be separately measured from downstream computational work.

ACA-C10The architecture can be subjected to bounded, falsifiable experiments comparing baseline computational work with ACA-gated computational work.

Explicit non-claims

ACA is intentionally bounded so experimentation cannot silently become marketing.

NON-CLAIM 1TA-14 ACA does not claim that every computation should be minimized, blocked, or reduced.

NON-CLAIM 2TA-14 ACA does not claim universal superiority over other computational architectures.

NON-CLAIM 3TA-14 ACA does not claim that every workload contains avoidable computation.

NON-CLAIM 4TA-14 ACA does not claim generalized proven reductions in electricity consumption, cooling demand, water consumption, emissions, infrastructure demand, latency, or economic cost.

NON-CLAIM 5TA-14 ACA does not claim that a blocked model call can be converted into a universal quantity of electricity, water, carbon, or infrastructure savings.

NON-CLAIM 6TA-14 ACA does not claim that computational admissibility automatically establishes authority for a later real-world execution.

NON-CLAIM 7Environmental and infrastructure benefits remain workload-specific and facility-specific hypotheses requiring measurement.

Computational accounting boundary

A governance gate does not earn an efficiency claim merely by blocking work. Its own cost must be counted.

Net computational benefit = Baseline total work − (ACA gate work + admitted downstream work + shifted/retry work)

GATE WORKRule checks, context fetch, policy lookup, classifier/model adjudication, and binding generation all count.

ADMITTED WORKModels, retrieval, tools, network, storage, and agent branches executed after commit count.

SHIFTED WORKRetries, deferred runs, work moved elsewhere, and escalation cost cannot be hidden.

VALID OUTCOMEAny positive resource result must preserve declared quality and acceptable false-decision rates.

TEST THE PROPOSITION

Measure it before claiming it.

The ACA Pilot Lab turns the registered proposition into an interactive accounting test. Compare baseline total work with ACA gate work, admitted downstream work, and shifted or retry work. Where measured data are available, extend the experiment to IT energy, facility effect, and water attribution without pretending a universal conversion exists.

Run ACA Pilot Lab →Open Pilot Kit →

Visible implementation: Admissible Search Gateway

The current public gateway is an early implementation surface, not a claim that the full upstream resource hypothesis has already been proven.

PROVIDER CANDIDATESA bounded provider-established candidate set is frozen with original rank preserved.

GOVERNED DELIVERYEach established candidate receives an ALLOW, HOLD, DENY, or ESCALATE determination before governed delivery.

ADMITTED EVIDENCEDelivered candidates remain tied to original provider rank, determination, reason, and bounded standing.

NON-ADMITTED EVIDENCEHOLD, DENY, and ESCALATE candidates remain preserved rather than silently erased or replaced.

Open the live gateway Measure the proposition Open ACA public area

ACA and Admissible Execution Architecture

Adjacent governance boundaries, not one collapsed architecture.

ACAGoverns whether computational work should be committed in its present form, scope, resource class, context, and intended consequence.

AEAGoverns whether a consequential action should be committed to execution in a real, organizational, financial, cyber-physical, or other externally consequential environment.

HANDOFFACA may produce evidence or candidate actions for AEA. The handoff transfers provenance and binding context, not execution authority.

BOUNDARYA computation may be admissible to produce a recommendation while the recommendation remains inadmissible to execute.

Current institutional status: TA-14 ACA Version 1.1 is registered as TA-14-AIGR-000039. Its governance identity is frozen and public. The resource proposition remains a falsifiable research proposition: if gate overhead equals or exceeds work avoided, quality degrades beyond the declared boundary, false decisions become unacceptable, or work is merely shifted, a compute-efficiency claim does not survive.

Next examination posture: founding demonstrations, adversarial examination, comparative architecture testing, interoperability examinations, and bounded pilots may now reference this registered baseline without silently modifying it.