What Makes an AI System an Agent, and What Does Not
Most things sold as AI agents are prompt chains with job titles. Here is the test that separates them, and why it changes cost, audit, and accountability.
Writing
Essays, analysis, and notes on applied AI, digital assets, investigations, and the systems around them.
- Sep 27, 2026 · AI Context Architecture: Designing AI Systems That Know What to Question A mature AI system knows which rules are fixed, which decisions are provisional, and when review is required. Context architecture makes that explicit.
- Sep 26, 2026 · Investigation A Case File a Reader Can Check A public case file should expose its evidence, limits, and unresolved questions so readers can inspect the reasoning, not just the verdict.
- Sep 26, 2026 · AI AI Agents Need Exploration Mode and Execution Mode Reliable execution needs stable boundaries. Useful research needs permission to challenge assumptions. One mode for both produces rigidity or risk.
- Sep 25, 2026 · Thoughts The Invisible Box: When AI Optimizes the Wrong System An AI agent can follow instructions perfectly and still move a business the wrong way. The failure may be the decision space humans created, not the model.
- Sep 24, 2026 · AI Guardrails Are Not Assumptions: A Context Hierarchy for AI Agents Security controls, architecture, and product hypotheses should not carry equal authority. A context hierarchy keeps boundaries firm and questions reviewable.
- Sep 23, 2026 · AI AGENTS.md Is Not Just Documentation Persistent agent instructions do more than describe a repository. They shape which options an AI weighs, what it avoids, and what becomes operating policy.
- Sep 23, 2026 · Investigation How an Investigation Ends Some investigations close, some become monitoring, some leave the chain entirely. Unsolved often means the answer sits with an institution or a court.
- Sep 22, 2026 · AI Awareness Before Architecture: The Hidden Risk in Agentic Systems Agentic systems can turn a vague idea into infrastructure before anyone tests it. Awareness before architecture stops fast commitment to the wrong thing.
- Sep 21, 2026 · AI The Work Behind AI That Looks Like Wasted Time Refining an AI workflow looks unproductive against one task. Across repeated work, context, checks, and judgment become an asset that makes output less fragile.
- Sep 19, 2026 · Investigation Shared Infrastructure Is Not Evidence of Common Control Two wallets do not share an operator because both touch a sponsor, bridge, router, or exchange. Widely used infrastructure is context, not proof of control.