Foundation
AI System Design
The foundational text — organizing intelligence around operational reasoning, ontology, logic mapping, decision rights, and human authority.
Read the primer →Resources
Resources
Essays on operational reasoning, AI system design, and building decisions that hold up where being wrong has consequences.
Demo
Judgment Call
A working demonstration of operational reasoning — set the situation and the priorities, and watch the system recommend, hand back, or escalate based on the consequence of error.
Run the demo →Diagnostic
Is There a System Around the Model?
A structured self-assessment for testing whether an AI initiative has the representation, logic, authority, and controls required for real work — run it against a stalled pilot or a build you are about to greenlight.
Run the diagnostic →Engagement
Discovery
How a first engagement works: reconstruct how the operation decides, make the logic explicit, test the evidence, and recommend what to build first, or whether to build at all.
Read about Discovery →November 2026
Decision Architecture for Operational Agents
How operational judgment is recovered from experienced people, proven against past decisions and given authority to act — the method and evidence behind our earlier essays, worked through one airline disruption.
Read the paper →October 2026
The Loop Is Not the System
The popular agent diagram explains how an agent runs, not how an operation decides — and why that gap stalls agent projects between pilot and operations.
Read the essay →October 2026
The Minimum Viable Pilot
Start with the decision, not the model — why the narrowest set of related decisions is the fastest, cheapest route from pilot to operations.
Read the essay →September 2026
How We Decide
How judgment moved from managers into the systems around them — and why agentic AI scales it only when the architecture preserves responsibility.
Read the essay →September 2026
Operational Reasoning by Design
A framework for enabling organizations to make better decisions — beginning with the operation, not the technology.
Read the essay →August 2026
Why Frontier Models Struggle with Complex Work
Frontier models are powerful cognitive components — and unreliable operating systems. Why complex work needs an architecture around the model, not a better prompt inside it.
Read the essay →Reading list
The AI Reading List, 2017–2026
A selection of the papers, reports and essays that have shaped, or are now shaping, current thinking about models, agents, reasoning, evaluation, governance, and the systems built around them.
Browse the list →