Research.
Independent projects from the practice.
What may act, what an adversary can learn, and what must keep working.
AuthorityThe Intent GapHow can consequential actions be constrained?Adaptive defenceProject RorschachWhat can an adversary learn, and at what cost?An UNMODELLED research programme.
Proposed research frameworkThe Convergence MapWhere authority, adaptive defence and resilience intersect.Operational resilienceOpen researchWhat must continue working under constraint?
Governance, assurance & communication
SagefolioWhat AI exists, who owns it, and is it approved?The Overlap MapWho owns the controls and evidence?BoardLevelWhat supports accountable sign-off?Seven TermsWhat must decision-makers understand?
Sagefolio provides the record of AI ownership, approvals and reviews. The Overlap Map examines accountability for controls, while BoardLevel supports decisions that require defensible sign-off.
- SCAMSENSECyber security for people.Before you click, pay or reply, ask Scout. Plain-English help with scams, for you, your kids and your parents.isitdodgy.org ↗
- SAGEFOLIOKnow your AI. Own the risk.An AI asset register and governance concept for organisations adopting AI. Record models, agents and AI-enabled tools, assign ownership, manage approvals and risks, and maintain a review history throughout their lifecycle.sagefolio.net ↗
- SEVEN TERMSExplaining AI to the people who sign off on it.A field guide for executives and boards: what each term does, what to ask before funding it, and where it creates security and privacy risk.View the guide →
- THE OVERLAP MAPWho owns what, once AI starts acting.An accountability framework for governance, security and compliance: who authorises action, operates controls and supplies evidence.View the map →
- BOARDLEVELApprove with defensible confidence.See what a material change touches and what remains unknown. Find the missing knowledge before accountable sign-off.boardlevel.net ↗
- THE INTENT GAPCan AI be trusted?A working paper on why authenticated identity no longer guarantees intent, and why trust has to move to verifying consequential actions.View the brief →
- THE CONVERGENCE MAPAuthority, adaptation, continuity.A proposed research framework connecting agent authority, adaptive defence and resilience under constrained power and connectivity.View the map →
- UNMODELLEDThe economics of autonomous attack.Concept-stage research into how cheaply agents can search, how long what they learn stays useful, and how to test the cost of disrupting that process.unmodelledai.com ↗
- PROJECT RORSCHACHWhat can an agent learn, and what does it cost?UNMODELLED’s experimental programme on controlled observation changes, reconnaissance accuracy and search cost. An illustrated introduction to the research.View the brief →