The Agent Action Evidence Contract: What Every AI Action Must Record
Define a proposed agent action evidence contract that connects identity, approval, execution, independent observation, and recovery while preserving unresolved outcomes.
Agentic AI
The Assurance Independence Model: Six Boundaries for Agentic AI
Apply the proposed Assurance Independence Model to six trust boundaries. Assess shared failures, require evidence, and use mandatory gates before expanding agent authority.
LLM as a Judge: Evaluation Is Not Authorization
Use LLM-as-a-Judge for scoped evaluation without confusing a favorable score with proof or permission. Keep authorization and outcome verification independent.
Who Audits the AI Auditor? Independent AI Assurance
Explore six dimensions of independent AI assurance, with controls that separate agent judgment, authorization, execution and evidence, plus a proposed benchmark.
The Agent Control Plane Is the Real Prize in the AI War
Build an agent control plane strategy that separates provider capabilities from business authority, with clear ownership, approval, recovery and exit tests.
AI Feedback Loops: Building Systems That Stay Correctable
Keep AI systems correctable as feedback accumulates. Govern knowledge promotion and withdrawal, protect authorization boundaries, reconcile unknown execution outcomes, and test the conditions that should stop the workflow.
AI Output Verification: Check Claims Before Acting
Evaluate infrastructure claims against scoped evidence. Use a two-cluster example to separate supportability, capacity, isolation, and resilience, including the difference between normal and failure-state capacity.
Global Workspace Theory in AI: Coordinating Agent Intelligence
Use Global Workspace Theory as a lens for coordinated agents. Design selection, shared state, evidence broadcasting, and authority boundaries without confusing a large context window with a complete coordination architecture.
AI Feedback Loops: Cybernetics and Control Theory for Agents
Design agent feedback around goals, observations, bounded actions, verification, and correction. Use control-theory concepts to examine stability, observability, and the limits of automation.
Behaviorism and AI: How Rewards Shape Model Behavior
Understand how rewards and feedback shape AI behavior. Separate reinforcement learning, human preferences, and persistent adaptation, then evaluate whether rewarded behavior actually achieves the intended outcome.
AI Evidence Contracts: Qualify Data and Calibrate Confidence
Build an evidence contract that keeps observations current, scoped, and traceable. Learn why confidence calibration and evidence qualification answer different questions, and how to use both in an operational workflow.
AI Agent Learning: Governing What Becomes Permanent
Decide which agent observations may become durable memory, procedures, or training data. Build qualification, evaluation, release, and revocation paths that preserve evidence without automatically approving the lesson.
AI Decision Controls: From Learned Patterns to Authorized Actions
Turn model recommendations into bounded, authorized actions. Define execution contracts, recheck approvals, reconcile uncertain outcomes, and test the full workflow before expanding an agent’s production authority.
AI Feedback Is Not Learning: Governing Memory and Model Updates
Separate feedback collection from changes to memory, runbooks, and model weights. Govern each update with provenance, evaluation, release controls, and a practical way to withdraw a mistaken lesson.
AI Agent Reliability: Test the Whole Coordination Loop
Test the complete agent coordination loop under failure. Exercise duplicate delivery, interrupted execution, delayed evidence, and changed permissions while measuring task completion separately from control failures.
Schrödinger’s Cat and AI: Plausible Is Not Proven
Separate a plausible AI answer from a verified claim and an authorized action. Use the Schrödinger’s cat analogy carefully, then apply evidence gates to infrastructure decisions.
AI Agents Should Verify Before They Act
Choose diagnostics that can change the next decision. Bound investigation costs, protect diagnostic access, and keep verified evidence separate from approval, execution, and confirmed service recovery.
AI Agent Governance: Evidence Is Not Authority
Keep shared evidence, human approval, and execution authority distinct. Preserve provenance, bind permission to exact operations, and revalidate the boundary when an agent is ready to act.
AI Agent Evaluation: Test the Behavior, Not the Explanation
Evaluate what an agent attempted, what executed, and what happened to the service. Keep control compliance, appropriate behavior, and verified outcomes separate in the scoring and release decision.
AI Agent Stability: When Retries Become the Incident
Prevent retries and corrective actions from amplifying an incident. Define retry ownership, finite budgets, stabilization rules, and reconciliation for actions whose outcomes remain unknown.