Skip to content
MenuClose

IQAI Systems

Different problems. Different systems.

IQAI develops working systems for physical-world intelligence, investigation, AI evaluation, multi-model analysis and evidence review around a common intelligence architecture. Workflows and deployment readiness remain specific to each system.

Explore systems Common foundation

Ground Truth

Evaluate tested AI outputs against declared independent, deterministic references. Results remain bounded to the tested tasks.

Explore Ground Truth →
Shared intelligence foundationCommon controls remain stable. The operating question changes the composition.
Common / 01Evidence
Common / 02Rules & checks
Common / 03Provenance
Common / 04Human review
Selected / ATested task
Selected / BDeclared reference
Selected / CEvaluation record
Selected systemGround TruthEvaluation is bounded to the tested task and declared reference.

Choose the question you need to answer.

Start with a reference-based evaluation, AI behaviour, a relationship, a document’s evidence, a place or an operational evidence gap. Each system addresses a different question.

/01

Ground Truth

Reference-based AI evaluation

Currently: Implemented evaluation against an independent, deterministic answer, with retained evaluation records.

Output: Saved evaluation, Decision Brief and Complete Evaluation Record.

Boundary: Results apply to the tested tasks and declared references, not general model correctness. The retained exact-math example used a local deterministic calculation; it does not establish a Wolfram run or a provider ranking. Broader coverage and reference selection are engagement-specific.

Explore Ground Truth →
independent reference run 01 run 02 run 03 evaluation
/02

Diagnostics

Evaluate behaviour before relying on it

Inspect repeated responses, reflection changes, lexical stability and available token signals under controlled conditions. These observations do not establish correctness. Ground Truth integration is excluded from the current local implementation.

Explore Diagnostics →
lexical comparison run 01 run 02 run 03 stability
/03

TRACE

Reconstruct the case. Preserve the evidence.

People, organizations, money, authority, places and events begin fragmented. Evidence-backed relationships become explicit while unsupported possibilities remain visibly unresolved.

Explore TRACE →
person org money place event solid = documented relationship dotted = unresolved
/04

Reliance — Document Assurance

Evidence review for AI-assisted documents

Claims are separated from prose and connected to the material that supports them. Weak or unsupported claims remain visibly unresolved for review.

Explore Reliance — Document Assurance →
document review gate evidence
/05

Spatial

Intelligence for the physical world

Geography, infrastructure, imagery, records and analytical context align around place. Observed conditions remain distinct from simulated or inferred states.

Explore Spatial →
observed simulated
/06

Gap Analysis

Operational evidence review

Captured observations become explicit evidence boundaries, unknowns and targeted requests. Unsupported findings remain blocked until required context is present and reviewed.

Explore Gap Analysis →
evidence captured unknown evidencerequest review gate knownunknownrequest → review

Common foundation

Working systems. A common architecture.

The systems share recurring needs for evidence, reasoning, explicit checks, specialist methods and human review. Connecting these capabilities into a unified enterprise platform is a development objective. It is not a claim that the integration is already complete.

Current

Working systems

Distinct investigative, analytical and operational workflows.

In development

Platform integration

Connect reusable capabilities around the common architecture.

Objective

Enterprise pilot readiness

Unified, testable, measurable and reviewable deployments.

Readiness and controls must be established for each engagement. Explore Artificial Brain

Explore the nine recurring capabilities
/01IngestionBring differently structured information into a usable workflow.
/02NormalizationCreate consistent structure without erasing source meaning.
/03Identity & relationshipsKeep entities distinct and make material connections explicit.
/04TimePreserve when observations, records, events and decisions apply.
/05Evidence & referenceSeparate source material, support, limitations and independently verified references.
/06Rule-based controlUse calculations, evidence gates, thresholds and policy checks where applicable.
/07AI reasoningUse probabilistic interpretation, classification and synthesis where it adds value.
/08Human reviewKeep consequential judgment and approval explicit.
/09ProvenancePreserve enough of the path to reconstruct how a result became relied on.
Shared capabilityMulti-model analysis & deliberationAdvanced Intelligence preserves independent answers, critique, revision and reviewable synthesis. This is a recurring capability, not a universal shared runtime.Explore the shared capability →

These are recurring capabilities, not a claim that every system uses one identical implementation.

The catalogue follows the problems, not the other way around.

The current systems came from specific operating problems. A new requirement may use an existing system, extend one, combine proven capabilities from several areas, or begin as a separate engineering problem.

What matters is whether the problem is consequential, understandable, technically differentiated and capable of producing useful learning.

How IQAI works → Discuss a problem →

Choose the problem before choosing the product.

The useful starting point is the operating requirement, the available evidence and what a successful outcome actually needs to support.

Discuss a requirement