Expound/Research
The papers, the articles, and what comes next
Two technical papers, the published articles, and the research program. Every result is stated with the campaign that produced it, so you can check it.
The technical record
The core argument, and the full source record.
Finality Assurance™: Core Technical Paper
The citable core argument: the problem, the formal result, the architecture, the benchmark evidence, the economics of Accepted Work, and the governed Work Management product.
Finality Assurance: Long Technical Paper
The full technical source record behind the core paper.
Published articles
Shorter arguments, all in one place.
Every published article and paper (on computing reliance, on what counts as work, on routing and execution, and on rights and economics) is collected on the Resources page, with the LinkedIn edition and the PDF wherever both exist.
Contributions to computer science
Five results the standard is built on, each stated with its boundary.
The core technical paper states five bounded contributions to computer science. Everything needed to evaluate them (definitions, theorem statements, assumptions, derivations and the empirical methods with their denominators) is in the papers. The later sections of the paper build, run and test them.
Conflation as a property of representation, not of practice
A formal result. Any status field smaller than a stated bound must merge at least one of the distinctions a downstream consumer is entitled to keep: permission (was it allowed), occurrence (did the effect happen), reliance (may someone act on it). And a field that can be set without regard to the truth destroys guaranteed information about ground truth at any vocabulary size. The deeper defect is settability, not vocabulary: conflation is a property of how state is represented and who can write it, not of how carefully a process is run. The result is about state representation in distributed systems and does not depend on artificial intelligence being involved.
A composition proof-obligation framework for decisions
Consequential decisions are composed from other decisions: in sequence, in parallel, contingently, fused, retried, compensated, superseded, retracted, looped. The work names the operators and the proof obligations each carries: safety preservation, liveness, identity, scope monotonicity, evidence provenance, failure propagation, cancellation, compensation, temporal validity, and equivalence to a reference semantics. Most systems compose decisions implicitly and cannot state what their composition preserves; the obligation set makes composition a checkable property rather than an emergent one.
Observation assurance is not consensus
A quorum proves that enough enrolled observers responded. It does not prove that the observer set was capable of detecting the event in question, and agreement among observers attached to the same incomplete boundary establishes nothing about what lay outside it. The missing predicate is observability rather than agreement, and the rule that follows is that not observed must not be converted into did not occur unless a calibrated observation topology licenses the inference. This is a distinct problem from Byzantine agreement.
An assurance measurement framework
Independent, complete, current, sufficient, material, adequate, strong authority, low risk: the load-bearing words of assurance are usually left undefined. The work requires each to be either measurable or explicitly policy-defined, and demonstrates one such measure in detail: independence expressed as a seven-dimension vector, with the rule that agreement among observers sharing a material common mode establishes none of its dimensions. That measure is testable in a way the word is not.
Explicit proof, observability, decidability and trust boundaries
The architecture states where proof ends and trust begins, including the recursive question of who verifies the verifier and where that regress is terminated by declaration rather than by proof. Rather than pretending to resolve undecidability, it governs what may be claimed and relied upon given it. Making that boundary explicit and machine-checkable is itself the contribution: a system that cannot say where its own proofs stop will eventually be asked to defend a claim it never had the standing to make.
Research program
What comes next.
Three lines of work extend the contributions above from the systems where they were established to systems the lab did not build.
The powered benchmark
A pre-registered study, with tasks spread across categories, on test environments built by other people, with wrong acceptances and wrong refusals reported as headline results beside cost.
Field validation
One consequential decision at a time, on a partner’s own systems against the partner’s own baseline, with the evidence retained under partner control.
Safety-critical and mission-critical domains
A validation agenda for clinical, transportation, energy, industrial and defense contexts: hazard analysis, integrity-level mapping, fail-safe behavior, human factors, regulatory engagement, so computed reliance earns its place beside the existing safety case.
Every result is stated with the campaign that produced it, so it can be reproduced on your own systems.
Next steps
Read it, then reproduce it on your own systems.
Every result is stated with the campaign that produced it. Independent validation partners are welcome.