Writing on applied AI, software engineering, and more.
A chronological collection of my writing on systems architecture, applied AI, the tech community, and everything in between.
October 19, 2026
Done Is a Claim: Designing Interfaces for Agentic Work
A task can outlast an agent’s reply. Its interface shows what remains provisional, what another system has confirmed and what the person can safely do next.
Read articleOctober 7, 2026
Backpressure is a Feature: Building Systems That Fail Gracefully Under Load
A queue can keep a system looking healthy while work grows too old to matter. Backpressure makes capacity visible early enough to refuse work deliberately and protect what remains useful.
Read articleSeptember 20, 2026
Building Tools for Agents: What Makes a Good Tool Interface?
Agent tools are interfaces for probabilistic callers. Good ones constrain interpretation, enforce authority and report outcomes clearly enough for agents and operators to act safely.
Read articleSeptember 15, 2026
The Agent Is Not the Architecture: Designing Reliable Systems Around LLM Agents
An LLM agent can decide correctly and still leave the system in the wrong state. Reliable architecture controls side effects and safely recovers from retries, restarts, delayed approvals, and uncertainty.
Read article