Writing

This is my blog.

AI, tech, and whatever else I'm thinking about.

  1. An Agentic Engineering Factory Inside GitHub

    GitHub already has the governance layer coding agents need. I use it to route agent work from a human-triaged issue through review and back to a human merge decision.

    AI Development Agents GitHub

  2. CRAFTS: The Agentic Coding Loop I Trust

    I wanted agents I could leave alone and still trust. CRAFTS is the coding loop I built to make autonomous output more reliable, more secure, and less forgetful.

    AI Development Agents

  3. Why I Put a Knowledge Base Inside My Codebase

    Spec files get stale. Agents work from outdated assumptions. Here's how an in-codebase knowledge base fixes that - and how the agents maintain it themselves.

    AI Development Agents

  4. TODO(HUMAN): The Case for Deliberate Presence in Agentic Development

    Full agentic autonomy is fast, but speed compounds in both directions. TODO(HUMAN) is the checkpoint system I use to stay present, sharp, and accountable in AI-assisted software development.

    AI Software Development Agentic Coding

  5. The n8n Rebuild Report: Real Numbers, Real Trade-offs, No Clean Ending.

    I predicted $44-70/month. It came in at $90. Here's the honest breakdown - cost, stability, creative quality, and the trade-offs that don't clean up nicely.

    AI Automation n8n Content Marketing

  6. I Spent $330 on an OpenClaw Content Agent. Here's What I Got.

    I built a 4-agent LinkedIn content pod on OpenClaw for an honest technical evaluation. Here's what worked, what broke, and why the cost math doesn't add up.

    AI Automation Content Marketing

  7. Why I Scrapped My OpenClaw Content Pod and Rebuilt It in n8n

    A spreadsheet audit of my Anthropic API costs revealed the real problem with agent-native automation. Here's what I rebuilt and the rule for evaluating AI agents.

    AI Automation n8n Content Marketing

  8. Building a Development Documentation MCP Server at Vimeo

    I built an MCP server that provides AI coding assistants with development documentation for Vimeo's custom codebase patterns, conventions, and libraries resulting in roughly 75% better output.

    AI MCP Case Study

  9. One User, One Banana: What My Failed AI Startup Taught Me About Validation

    I spent 5 months building an AI startup, got 50 signups, and watched my only active user upload a banana. Here's what it taught me about enthusiasm vs. actual demand.

    AI Case Study
Subscribe
Get new posts when they go up. No newsletter, just RSS.
RSS feed Email me instead