Writing about AI agents, engineering workflows, and building things that actually work.
Start here
What Nobody Tells You About Running MCP Servers in Production
Six months, three production MCP servers, twelve real failures. The thing that broke everything wasn't the protocol — it was the descriptions.
Read the recommended first post ->
agent security: what microsoft shipped (and what you still need to build)
May 26, 2026Microsoft Entra Agent ID provides identity and OAuth. It does not prevent jailbreaks, enforce tool boundaries, or stop data exfiltration. Here is the architecture gap and what to build.
How I Actually Test AI Agents
May 25, 2026Everyone talks about evals but nobody shows you the boring part: the labeled test set, the scoring script, and what you do when the number doesn't move.
What Nobody Tells You About Running MCP Servers in Production
May 23, 2026Six months, three production MCP servers, twelve real failures. The thing that broke everything wasn't the protocol — it was the descriptions.
How to Write a Good Skill for an AI Agent
May 12, 2026Skills are the difference between an agent that actually works and one that confidently does the wrong thing. Here's what I learned building 50+ of them.
Writing Good AI Agents: What I Learned Running a Fleet of Them
May 8, 2026Practical lessons from deploying and running multiple AI agents in production — what actually works and what blows up.
Building Better AI Agents: Lessons from Running 40+ in Production
May 7, 2026Why constraint, clarity, and oversight matter more than intelligence when working with AI agents.
From Copy-Pasting ChatGPT to AI That Runs My Workflow
May 5, 2026How I went from pasting errors into ChatGPT to having Copilot CLI fix tests, create PRs, and manage my projects.
First Post (Again)
May 4, 2025Starting a blog about AI agents. Again. This time might actually stick.