Insights

Field notes from the work, not thought-leadership for its own sake.

Guarding the Way: How to Actually Evaluate an AI Agent Runtime
You cannot evaluate an agent runtime by reading its architecture docs. A security invariant in a comment is a request. In the type system, it is a fact.
Brick for Stone: The Limits of Specifying Everything Up Front
Spec-driven development is genuinely useful. The trouble starts when a capable AI agent makes it easy to specify the wrong thing very efficiently.
AIpocalypse, Just-in-Time
The AI doom discourse has detached from the experience of using the tools. The hype follows an exponential curve. Real workflows follow a linear one.
Software for Strangers
Distance is not just geographic, it is relational. The systems nobody watches break down and stay broken, because correctness itself is relational.
Framing Freedom: A Governance Model for Agentic AI
Most AI failures are not about too much control or too little. They share one root: nobody understood that constraint is what gives agency its shape.