AiNOS
§T01 · CORE TECHNOLOGY

Runtime specimen · accountable AI

Core Technology

AiNOS compiles source signals into an ontology substrate, routes each judgment by type, and leaves every decision as an inspectable trace.

deterministic routesemantic reasoning
01source signals02ontology03router core04trace ledger
A runtime cutaway showing source signals flowing into an ontology substrate, passing through a router core, and becoming signed rows in a trace ledger.01 SOURCE SIGNALS02 ONTOLOGY SUBSTRATE03 ROUTER CORE04 TRACE LEDGERDOCSAUDIOVIDEOOPStyped domain substratetype routertrace ledgerROUTEDET · RULE 7.3GROUNDONTOLOGY §4.2TRACE#T01-A114
FIG. T01 — RUNTIME CUTAWAY: SOURCE SIGNALS COMPILED INTO AN ONTOLOGY SUBSTRATE, ROUTED THROUGH THE CORE, AND RECORDED IN A TRACE LEDGER · DET · SEM
§T02 · COMPILER

DOMAIN REALITY COMPILER

Signals become typed operating structure.

AiNOS registers domain evidence into ontology objects: square deterministic nodes for knowable structure, round semantic nodes for interpreted meaning, and typed edges between them.

DOCOPSSENSOR / VIDEOCORRECTION
Registration
Every source enters with class, timestamp, actor, and observed object.
Typed output
Nodes and edges preserve what is deterministic versus what was inferred.
Documents, operational events, sensor video, and human corrections pass through registration gates and become typed ontology nodes and edges.DOCOPSSENSORCORRECTALIGN · NORMALIZE · TYPEROLEASSETINTENTEVENTEXCPOLICY
FIG. T02 — SOURCE SIGNALS REGISTERED INTO TYPED ONTOLOGY STRUCTURE · DET · SEM
§T03 · ROUTER

SELECTIVE DETERMINISM

Judgment routes by type.

AiNOS does not ask one model to handle every decision. It separates what can be known structurally from what must be interpreted against grounded evidence.

STRUCTURE

Knowable structure

Schemas, state transitions, thresholds, graph constraints, and audit rules run as deterministic logic.

INTERPRETATION

Grounded interpretation

Ambiguous signals route to semantic reasoning constrained by ontology, source evidence, and trace requirements.

A judgment token enters a runtime router, then splits into a square sky deterministic route for knowable structure and a round violet semantic route for grounded interpretation.JUDGMENTSTRUCTURErule / graph / stateINTERPRETATIONevidence / contextROUTETRACE
FIG. T03 — JUDGMENT ROUTED BY TYPE · DET · SEMROUTE / TRACE
§T04 · TRACE

Decision trace specimen

One release decision, inspectable end to end.

AiNOS keeps the chain of custody intact: evidence, entity binding, route type, grounds, judgment, and signature.

constraintsgrounded exception
Case

Port berth release readiness, anonymized case PR-042

Decision

Hold release. Review again at 19:45.

Why

Yard blockage and AMR queue fail deterministic limits; gust advisory adds exposure risk.

Trace

Signed evidence, entity bindings, route type, grounds, and judgment.

The proof is not that a model answered. It is that every step of the answer is typed, grounded, and replayable.

An anonymized port berth release readiness decision trace showing evidence intake, ontology entities, route split, deterministic and semantic grounds, final hold decision, and signed audit hash.TRACE SPECIMEN / PORT RELEASE READINESS / CASE PR-042STATIC AUDIT VIEW01 EVIDENCE INTAKEE01 BERTH 14WINDOW 19:20-21:00E02 YARD A3BLOCKAGE 31%E03 AMR FLEETQUEUE 17 MINE04 NOTICEGUST ADVISORY02 ONTOLOGY ENTITIESVESSELANON-42BERTHB14YARDA3AMRQ-17WEATHERGUSTROUTEBY TYPECONSTRAINTSberth window okyard blockage failAMR queue limitEXCEPTIONgust notice raisestug exposure risk05 FINAL JUDGMENTHOLD RELEASE / REVIEW AT 19:4506 SIGNED AUDIT TRACESIG 7F4C-A19E / ONT v3.8AUDIT SIGNEDCOPY IS ANONYMIZED. THE TRACE PRESERVES INPUTS, ENTITY BINDINGS, ROUTE TYPE, GROUNDS, JUDGMENT, AND SIGNATURE.
FIG. T04 - BERTH RELEASE READINESS TRACE DET SEM
§T05 · LEARNING
ontology learning loop

Exceptions become edges, not anecdotes.

Every exception, correction, and accepted override is stored as a trace that can change the model: denser ontology edges, sharper rule boundaries, and richer scenario memory.

rule boundaryscenario memory
Exception and correction traces convert a sparse ontology into a denser model with sharper rule boundaries, higher edge confidence, and richer scenario memory.BEFORE · SPARSE OPERATING MODELAFTER · CORRECTED OPERATING MODELmissing exception edgeTRACEexceptioncorrectionRULEBOUNDARYSHARPedge confidencescenario memory
FIG. T05 — CORRECTION TRACES DENSIFY THE MODEL · DET · SEMFINAL STATE BY DEFAULT

  1. Exception trace

    A failed route is kept as a typed contradiction: the missing edge, the violated rule, and the scenario that exposed it.

  2. Correction trace

    Human correction writes back as ontology structure: new edge, adjusted confidence, sharper boundary condition.

  3. Scenario memory

    The resolved case remains addressable, so future judgments compare against lived operating history instead of a generic prompt.