Architecting Authority in Intelligent Systems
Trust Is Infrastructure
An agent sends what it was told only to review. Separate the reason for the action from the right to take it.
Two Proofs of Action
Enter this worldDecisions beneath the system
Audit authority. Map the write path. Exercise the exit. Evaluate the action. Map the decision. These are the five moves at the heart of the collection—and five ways to see what an intelligent system must preserve when it begins to interpret, remember, choose, and act.
Start with the decision your system is already making.Each talk starts with a recognizable failure: an invoice sent after “do not send,” a deleted address that still directs a shipment, model weights without an operable system, one Arabic intent that produces different actions, or a workflow whose next step is no longer fully specified in advance.
The scene comes first; the machinery follows. Intent, state, authority, dependency, representation, and decision become visible so that leaders, practitioners, and researchers can leave with a choice to examine rather than a trend to admire.
Central question
An agent sends what it was told only to review. Separate the reason for the action from the right to take it.
Architecting Authority in Intelligent Systems
An agent sends what it was told only to review. Separate the reason for the action from the right to take it.
Two Proofs of Action
Enter this worldArchitecting State and Inference Across Time
A deleted address still directs a shipment. Trace the derived state, then decide when the past must lose its power to act.
Effective Forgetting
Enter this worldArchitecting AI Capability
Own the weights, then test the system. Capability is what remains when a supplier, region, or operating assumption changes.
Capability Under Stress
Enter this worldFrom Expression to Action in Intelligent Systems
One intent arrives in five forms. Judge the system by the action it takes—and by whether it knows when to ask.
Intent-to-Action Fidelity
Enter this worldArchitecting Delegation and Decision Boundaries
Before deploying an agent, map which decisions are fixed at design time and which may move into runtime.
The Specification Frontier
Enter this worldEvery talk has a stable root, a moving frontier, and a room that makes the route between them new each time.
The question and signature idea that give the talk its identity and remain clear as the world around them changes.
New research, systems, and institutional realities—evidence, cases, and questions that extend the talk without replacing its core.
The audience’s experience and context, shaping which doors open, which examples come closer, and which question deserves more time.
A shared understanding of the central question for leaders and mixed audiences.
The relationships and trade-offs that practitioners and decision-makers need to examine.
For engineers and researchers, with prerequisites and interaction agreed in advance.
These formats are proposals. Readiness, duration and interaction are confirmed for each invitation. These are the languages of the website descriptions. Live delivery languages and interpretation arrangements are confirmed separately for each engagement.
They are technical talks with human consequences. Each begins with a scene people can feel, opens the machinery until the causal structure becomes visible, and returns to the institution with a clearer decision.
Everyone encounters the same story at a different resolution: a mental model for the newcomer, a mechanism for the engineer, a hypothesis for the researcher, and a decision for the leader.
Each talk has a stable root, a moving research frontier, and a different room. It does not change merely to display new headlines. It changes when the enduring question meets new evidence and a new audience.
One image, one proposed signature idea, one question that resists easy closure, and one action to try: audit authority, map the write path, exercise the exit, evaluate the action, or map the decision.
The depth and examples can respond to the audience while the underlying meaning, evidence and limits remain fixed.
Public pages describe scope, audience and intended learning outcomes. Participant materials are discussed separately.
No. Readiness, format, language and date require separate confirmation.
A talk is a bounded intellectual contribution. Training or implementation work needs a separate scope and set of outcomes.
The title and central question return, but the route changes with the evidence and the people in the room. Each edition ends in a decision the audience can inspect at work the next day.
Five questions beyond the tool
Choose the closest question, then carry that context into the speaking page for a more precise conversation.
Discuss a talk