September 16, 2026
The Aquarius Reset
The origin story brings together missed calls, illness, recovery, invention, and the author's interpretation of their timing.
Read moreHome / Publications
Publications
Essays from the field.
Authority, coordination, and the gap between what a system can do and what it was authorized to do.
Explore topics & reading paths
45 published essays · newest first · full text and source PDFs
Visit the Substack archive
September 3, 2026
Ham Sandwich
Buying a sandwich sounds simple until an agent must honor the buyer's price, ingredients, merchant, and spending limits throughout the purchase.
Read more
September 2, 2026
The Drone on a String
A reliable control link does not establish who may authorize the next consequential action.
Read more
August 30, 2026
The Agent Is Still an Agent
An agent can plan and act without becoming the principal who authorizes the work.
Read more
August 18, 2026
Is It AI Infrastructure Week Again?
Debates about AI infrastructure often begin with how much capacity to build.
Read more
August 15, 2026
Every Hop a Signature
A signature can establish a delegation step without settling how much repeated verification a long chain should require.
Read more
August 10, 2026
The Agent at the Gate
Admission to a website or platform does not settle an agent's authority to act there.
Read more
August 6, 2026
You Can Feel It. You Can’t Name It Yet
A promise, consent, or permission can lose its force at a handoff even while each system reports success.
Read more
August 5, 2026
Your FTC Settlement Didn’t Close Your Exposure
A shopper can move through an ad, chat, calculator, lender, and contract while the authority behind a representation falls out at each handoff.
Read more
July 25, 2026
AI Is the Canary
AI makes an older systems problem harder to ignore: people once carried permissions and context across gaps that automation now crosses on its own.
Read more
July 20, 2026
False Refusal
A system can fail by refusing legitimate work as well as by permitting an unauthorized act.
Read more
July 16, 2026
Google, the EU, and the Right to Act
Cross-platform access raises a second question: what may an admitted agent actually do, for whom, and for how long? This essay reads interoperability policy through that operational distinction.
Read more
July 15, 2026
Anthropic Claude's Holy Grail
This satirical piece imagines an AI evaluation system that can change the conditions of its own success.
Read more
July 9, 2026
A Note from the Inventor of Hermes-Echo
The inventor's commercial introduction explains how the portfolio grew from specific communication problems into a broader argument about governing live interactions.
Read more
July 8, 2026
The Missing Dimension
Identity, routing, memory, payment, and accessibility each describe a function.
Read more
July 7, 2026
Before the Market Names the Primitive
What makes an architectural primitive different from a collection of features? This essay locates SSOAR in a constraint: a live session may change while its governing authority remains continuous, without creating a second independent authority root.
Read more
July 1, 2026
Schrems III and the Sea
Using a literary analogy, this essay questions whether repeated institutional assurances can resolve a cross-border data-governance problem.
Read more
June 25, 2026
The Fifth Ledger
If compute, billing, entitlement, and authority can describe the same bounded undertaking, energy attribution becomes another question to examine.
Read more
June 25, 2026
The Floor Nobody Measures
Efficiency measured inside a model or controlled experiment can leave coordination across providers and authority domains outside the accounting.
Read more
June 12, 2026
Ownership Cycles
An invention's reach and the structure that owns or licenses it are different questions.
Read moreJune 4, 2026
IoE Impasse
Connecting more things does not by itself settle their identity, authority, or obligations when they act together.
Read more
May 25, 2026
The Token Booth
Selling compute capacity and governing an authorized interaction describe different commercial positions.
Read more
May 22, 2026
The Four Token Ledgers
Cheaper inference and larger AI bills can coexist because they describe different quantities.
Read more
May 22, 2026
Tokens and Authority
When inference is sold through a subscriber plan, payment, entitlement, and permission to participate can be mistaken for the same thing.
Read more
May 16, 2026
Five Walls, One Perimeter
Five different institutional approaches to AI can encounter the same control problem.
Read more
May 14, 2026
Slip Sliding Away
Storage, retention, memory, and focus are not interchangeable.
Read more
May 14, 2026
You Cannot Cliff Note Sherlock Holmes
A summary may retain a conclusion while losing the exceptions and evidence that made it authoritative.
Read more
May 10, 2026
Component Truth Is Not Interaction Truth
Healthy components do not guarantee a successful interaction.
Read more
May 10, 2026
Contingent Accessibility
An accessibility feature can exist without delivering access to the person who needs it.
Read more
May 8, 2026
Outsourced by Accident
A vendor can acquire practical control over an institution's actions without receiving an explicit mandate for every consequence.
Read more
May 7, 2026
Compute on the Side of the House
Moving compute onto a house changes more than its location.
Read more
April 30, 2026
The Illusion of Autonomy
An agent's apology after a destructive action does not supply the authority boundary that should have governed the action.
Read more
April 22, 2026
Distributed Systems in a Palliative State
The medical analogy here describes systems kept running through repeated interventions while the underlying coordination problem persists.
Read more
April 21, 2026
Concurrence Collision
Concurrency inside one controlled system differs from multiple control domains changing at once.
Read more
April 20, 2026
The Détente of Non-Compliance
Organizations often rely on policies, attestations, and retrospective audits where interaction-level proof is difficult to obtain.
Read more
April 18, 2026
The Coordination Limit, Part II
The second coordination essay asks whether optimizing individual components changes the structure of the total coordination burden.
Read more
April 12, 2026
Glasswing: AI Grading Its Own Homework
Security findings are harder to trust when the system producing them can also influence their evaluation or supporting record.
Read more
April 10, 2026
AI Is Failing, But Not for the Reason People Think
Retries, context loss, and broken workflows can arise even when a model answers well.
Read more
April 9, 2026
The Wrong Architecture Always Escapes
Containment becomes harder to assess when the governed system can influence its own audit record or evaluation.
Read more
April 4, 2026
AI Hot Rod
An engine's test-bench performance does not describe the whole car on the road.
Read moreApril 2, 2026
Just Another Lab Rat
This personal account describes AI as cognitive assistance while the author retains judgment and responsibility.
Read more
April 1, 2026
The Coordination Limit
How much work does a system spend coordinating the work it was built to do? This essay proposes a model in which interacting participants, modalities, features, authorities, and transports increase that burden.
Read more