Notes
Field notes from product work, workflow systems, agent tooling, architecture decisions, and the debugging behind useful builds.
Expensive Tokens Reward Good Architecture
As AI token subsidies fade, engineers with modular systems, fast feedback loops, and swappable model components will have a cost advantage.
Don't Let the LLM Decide What a Word Is
A reading app for Spanish learners taught me which decisions belong to an LLM and which belong in code. The boundary is sharper than I thought.
Stop Reviewing AI Code. Give It Tools to Verify Its Own Work.
The shift from reviewing AI-generated code to giving AI agents the feedback loops they need to catch their own mistakes.
Browser Automation for AI Agents: Playwriter vs Agent-Browser vs Claude in Chrome
A hands-on comparison of three browser automation tools for AI agents. Benchmarks, code examples, and recommendations for different use cases.
Choosing a Backend in 2026: Convex, InsForge, or Supabase?
A practical comparison of three backend platforms for indie devs and startup CTOs. When to use each, code examples, and my honest take after building with Supabase.
I Hate Creating Content, So I Built a Robot to Do It For Me
How I automated my own content workflow with a functional AI pipeline that turns call transcripts into ready-to-schedule posts.