STATE · CONTEXT · TIME
Continuous infrastructure reasoning
How can a model maintain a faithful representation of assets, identities, dependencies, changes, and risk while the environment continues to move?
GLEMAD RESEARCH
Reasoning under pressure, uncertainty, and consequence.
We study the conditions under which intelligent systems can contribute to infrastructure defense without losing contact with evidence, authority, or real-world consequence.
Enter the research indexTHE PREMISE
Infrastructure defense is not a classification problem. It is a continuous problem of state, interpretation, and judgment.
Signals arrive from systems that were not designed to tell one story. Attackers exploit the gaps between those systems, the delay between observation and understanding, and the uncertainty that surrounds consequential response.
Our research asks what intelligence must become when the environment is live, the evidence is contested, and every action changes the conditions being studied.
ACTIVE RESEARCH AGENDA
The questions that organize our work.
HYPOTHESIS · CONFLICT · CONFIDENCE
Threat interpretation under uncertainty
How should a system distinguish legitimate change, operational drift, misconfiguration, and malicious intent when the evidence remains incomplete?
AUTHORITY · CONSEQUENCE · REVERSIBILITY
Controlled defensive action
What constraints must govern assistance, containment, escalation, and execution when an incorrect action can disrupt essential services?
EVALUATION · RED TEAMING · FAILURE
Evaluation and adversarial resilience
How do we measure reasoning integrity when the system is exposed to deception, compromised inputs, coordination failure, and novel attack behavior?
OUR METHOD
A claim becomes useful when it can survive challenge.
We structure research around observable state, explicit uncertainty, adversarial testing, constrained decisions, and records that allow others to inspect how a conclusion was reached.
Observe
Build time-aware records of infrastructure state with source and ownership intact.
Challenge
Test competing explanations instead of rewarding the first plausible answer.
Constrain
Evaluate decisions against permission, policy, consequence, and reversibility.
Verify
Measure whether an intervention changed both the threat and the operating state.
Preserve
Keep evidence connected from first observation through final outcome.
CURRENT MODEL
Ollandi 5 Preview.
Ollandi 5 Preview is the latest model in the Ollandi model class. It advances Glemad’s work in persistent infrastructure state, cross-domain threat reasoning, constraint-aware response, and evidence preservation.
Earlier model research remains part of the published record. Current development is organized within the Ollandi model class.
Read the technical paperTHE RESEARCH RECORD
Methods, findings, limitations, and revisions belong in public view.
RESEARCH COLLABORATION