When the Instrument Is the Lie: Why Dashboards Cannot Replace Judgment
Buncefield, Milford Haven and Texas City all share one detail boards forget: the reading itself was wrong. A dashboard culture assumes the gauge tells the truth.
The journal
Notes between covers.
The argument is long-form and lives in the book. The questions it raises arrive weekly, and they are shorter — working notes from the writing, and from the rooms where boards are already asking them. Signed pieces carry their author’s name; unsigned notes are the house’s, held to the same standard.
Have them as they are written
One email an entry. The publication date goes to this list first.
October 2026
1 entry
September 2026
4 entries
PwC Gives Boards Six Actions for AI. None of Them Is Judgment.
The board guidance is structurally sound — allocate oversight, assign accountability, monitor outcomes. But it assumes boards can evaluate management's AI judgment, and says nothing about how they acquire their own.
A Policy Layer That Approves Everything Is Not Oversight
The proposed execution layer between AI agents and on-chain transactions answers the wrong half of the delegation question.
Your Agent's Loyalty Is a Design Choice, Not a Default
The duty-of-loyalty framing for AI agents is right — but detection, not liability, is where delegation actually breaks.
Master, worker, operator.
The figure is a single line drawn through a century of work, and every step on it was a good idea.
August 2026
9 entries
The bench can be the same size and still form nobody.
The headcount evidence about AI and early-career work is contested in both directions. The question that holds is what a junior actually does in a day.
Two kinds of knowing.
An objection to our own argument that we think is right, and the test the framework owes because of it.
Did the work come back marked.
A randomised trial with nearly a thousand students shows what separates work that built judgment from work that only produced an answer, and the test it yields fits on a single memo.
Argyris put the failure in the room.
An essay on skilled incompetence sent us back to the 1986 original, where the failure sits between people — which is the version a board can do something about.
Can Boards Govern AI in Real Time?
Addressing the challenges and capabilities needed for real-time AI governance.
Judgment Capital — the book, and what it argues
The launch post: what judgment capital is, why it goes unnoticed, and where the board paper lives.
Human Oversight in AI: A Business Imperative or a Competence Drain?
Explore the tension between AI oversight and competence.
What this journal is.
Working notes from the writing of Judgment Capital, and from the rooms where its questions are being asked.
AI Governance: A Fiduciary Duty for Boards?
Exploring the board's role in AI governance and fiduciary duties.