Writing

Notes on making AI useful.

Essays and case notes about governed agents, operating systems for AI work, small products, and the practical decisions that matter after the demo. Every piece is grounded in a system I have built, directed, or am actively designing.

Essays & field notes

Seven pieces on architecture, operations, product judgment, and bounded automation.

01 / Operating architecture

I Built an Operating System for Governed AI Work

A practical architecture for persistent orchestration, specialist workers, deterministic controls, and human authority over consequential action.

8 min read
02 / AI governance

Why I Configure My AI Agents to Challenge Me

Designed dissent, calibrated friction, and the authority boundaries that keep a helpful agent from becoming either an agreement engine or a synthetic manager.

8 min read
03 / Systems operations

Four Surfaces, One Operating Record

How Hermes, Vikunja, Discord, and Docmost divide execution, commitments, conversation, and durable knowledge without multiplying sources of truth.

7 min read
04 / Product architecture

The Film Should Be the Interface

A provider-neutral production architecture built around durable assets, shot intent, visible worker passes, and explicit generation and publishing gates.

7 min read
05 / Product direction

Drift Tones: Making a Small Browser Tool Feel Complete

From technical spike to restrained browser product: Web Audio, accessible state, safer language, and a tool-first redesign after rejecting the first polished pass.

6 min read
06 / Humane tooling

Soft Deadline Room: designing a timer that does not shout

A private, dependency-free timer prototype that explores calm interaction design, local-only behavior, accessible motion, and honest claims.

5 min read
07 / Agent architecture

I built an autonomous AI agent for almost nothing—and it started asking its own questions

Why Python should do the lifting, why small models need bounded decisions, and what happened when V’ger was allowed to wander.

8 min read