Blog posts

Discovering Claude Code's advanced features beyond basic prompting: hooks, skills, subagents, and custom workflows for technical writing.

How code execution with MCP reduces token usage by giving AI agents a workspace outside the context window for data-heavy processing tasks.

Reflections on AI-assisted coding, personal agency, and why working at higher levels of abstraction doesn't mean losing mastery or understanding.

Building an AI agent to test documentation information architecture by simulating how users navigate docs to find answers.

Using the Claude Agent SDK to play the Wikipedia game as an experiment in testing information architecture with AI.

Why enthusiasm for AI tools isn't always contagious, and what I learned about sharing new technologies with skeptical colleagues.

Why documentation metrics are harder to interpret than SRE dashboards, and how to think about measuring docs impact without fooling yourself.

How creating onboarding documentation for a new teammate also improved my AI tools by providing the same context engineering both need to succeed.

A critical look at the MIT 'Your Brain on ChatGPT' study and why sensationalized AI research headlines make productive conversations harder.

A hands-on guide to connecting MCP servers to AI clients, debugging with curl, and building a multi-source documentation research agent.

Why MCP exists and how it helps LLMs use APIs more reliably by providing structured guardrails instead of relying on hallucinated parameters.

A detailed post-mortem of a Kubernetes outage where a routine video processing job brought down an entire streaming cluster for five days.