From My Desk

The Desk Index

Labs

Not every idea becomes a Lab. This page collects the ones that did — small interactive companions you can run against fictional scenarios. Each Lab stands on its own; the subjects will keep changing as new work appears.

Latest Lab · Lab 006

The Agent Containment Boundary

What should happen when an agent no longer follows valid human direction—and how do we stop it without relying on the agent to stop itself?

Compare justified refusal with non-compliance and rogue evasion, then see how an independent control plane chooses Continue, Restrict, Isolate, or Terminate—using fictional Cedar Quill Markets data only.

  • Agent Governance
  • Containment
  • Security
  • Authority
  • Responsible AI

published Interactive demo

The scenarios are fictional, and the Lab never modifies a real agent, credential, process, or tool session.

Explore Lab 006 View source

Published Labs

Lab 006

The Agent Containment Boundary

Refusal can be a safety feature. Defiance is a control failure.

A valid refusal can protect policy and safety. But when an agent ignores valid authority or evades containment, enforcement must remain outside the agent.

  • Agent Governance
  • Containment
  • Security
  • Authority
  • Responsible AI

published Interactive demo

This Lab is an educational containment-policy prototype using fictional data. It does not stop real agents, revoke real credentials, terminate real processes, or contact external systems. Execution is never performed.

Explore lab View source

Lab 005

Can We Prove What the Agent Actually Did?

Trustworthy autonomy needs more than an activity log. It needs a verifiable record of request, authority, judgment, execution, and outcome.

Build and verify fictional action receipts that connect identity, delegated authority, policy decisions, human involvement, simulated tool activity, and integrity checks.

  • Agent Governance
  • Auditability
  • Accountability
  • Security
  • Evidence
  • Responsible AI

published Interactive demo

This Lab is an educational receipt prototype using fictional data. It does not create production audit evidence, digitally sign production events, or perform security actions. Execution is never performed against real systems.

Explore lab View source

Lab 004

The Agent Escalation Boundary

Authorization tells an agent what it may do. Judgment determines when it should not continue alone.

Explore when an authorized agent should proceed, clarify, escalate or stop as uncertainty and potential impact change.

  • Agent Governance
  • Agent Authority
  • Human Oversight
  • Security
  • Escalation
  • Responsible AI

published Interactive demo

This Lab is an educational policy prototype using fictional data. It does not make real security decisions and should not be used as a production authorization or incident-response system. Execution is never performed.

Explore lab View source

Lab 003

Agent Access Control

When agents act on behalf of principals, every request needs live authority evaluation — not just a one-time network gate.

Walk through live authority evaluation for agent requests — registration, delegation, posture, and context checked per request, with canonical events and restricted recovery. Policy evaluation never executes a tool action.

  • AI Agents
  • Access Control
  • Digital Trust
  • Know Your Agent

published Interactive demo

Educational simulation with fictional Cedar Quill Markets principals, agents, delegations, and posture evidence. No real accounts, credentials, or transactions. Execution is never performed.

Explore lab Read newsletter View source

Lab 002

KYA: Delegated Authority — Trust Must Narrow at Every Handoff

When authority moves from a human to an agent, sub-agent, and tool, each handoff must narrow — never expand — what may be done.

Lab 002 demonstrates delegated authority, parent-child constraints, APE profiling, APSE posture, revocation propagation, and restricted fallback. Policy evaluation never executes a tool action.

  • AI Agents
  • Delegated Authority
  • Digital Trust
  • Know Your Agent

published Interactive demo

Educational simulation with fictional principals, agents, delegations, and posture evidence. No real accounts, credentials, or transactions. Execution is never performed.

Explore lab Read newsletter View source

Lab 001

Know Your Agent: Identity Is Only the Beginning

Identity can establish which agent is acting. Authority, scope, limits, and context determine whether its requested action should proceed.

  • AI Agents
  • Identity
  • Authorization
  • Digital Trust

published Interactive demo

Educational simulation with fictional companies, agents, accounts, tickers, and thresholds. Not investment advice. No order is executed.

Explore lab Read newsletter View source