LPS + DTII
Continuously re-orders competing goals while preserving task identity under change.
Independent R&D · Toronto, Canada
Intelligence,
engineered for trust.
Extrados Intelligence develops formally grounded, verifiable architectures for autonomous systems—designed for environments where performance alone is not enough.
Formal methodsAutonomous systemsVerification & compliance
01 / Mandate
Autonomy should be auditable before deployment, not merely explainable after failure.
Our work focuses on deterministic reasoning, verifiable internal state, and explicit conflict arbitration—building toward autonomous systems that can be certified and trusted in mission-critical contexts.
02 / Foundational research
Submitted June 2026
Published research
A Formal Substrate for Artificial General Intelligence
Matthew McLeod · Independent Researcher
Toronto, Ontario, Canada
A non-LLM, non-reinforcement-learning architecture for general autonomous agency—built on deterministic reasoning and verifiable state rather than statistical inference alone.
State Verification Principle
Requires internal system state to be verified, rather than assumed, at every decision point.
ADL / CAL
Detects and arbitrates conflicts between competing values or objectives before action.
The Norman Test
An adversarial pre-deployment certification protocol, analogous in intent to IEC 61508.
03 / Defence innovation
Applied systems
Proposed concept
Fusion Compliance Engine (FCE)
A deterministic, policy-aware compliance and enforcement layer for multi-sensor AI fusion systems—intended to help ensure classification-sensitive outputs remain auditable and standards-compliant in operational environments.
Addressing the S&T challenge: “Reliable AI Sensor Fusion for Real-World Missions.”Filed under With A Thought, Matthew McLeod's prior research entity. The concept is under review and is not confirmed or funded. Program, contract, and technical details are withheld pending outcome.
04 / About
Founder profile
Matthew McLeod
Building autonomy that can be certified and trusted, not just benchmarked.
Matthew works independently on formal, verifiable architectures for autonomous systems—spanning theoretical work on general intelligence and applied engineering for defence-relevant sensor fusion and compliance systems.
Founder, Extrados Intelligence Inc.
Independent Researcher · Toronto, Canada
05 / Contact
For serious problems,
serious collaborators.
Open to research collaboration, employment, and program inquiries.