A few of the pieces most relevant to the work are below. The full archive is on Medium.
Markdown became where I curate context, directives, and memory for the model. HTML became how I show finished work to people. Why each format won without anyone choosing it, why neither one shares cleanly, and the two things I built to close that gap.
Read Markdown for the Machine, HTML for the Humans on Medium
Then I had to sleep at night. The containerized setup I built so a coding agent can run wide open without touching anything it shouldn't.
Read Containerized Development: My Security Layer for AI Coding Tools on Medium · Code on GitHub
It replaced years of my reflexes with grep. The case that agentic search is really reranking in disguise — and where embeddings, RAG, and vector databases still earn their place.
Read Your Coding Agent's grep Loop Is Just Reranking in Disguise on Medium
SQLite, real invoice PDFs, all the S-Corp money plumbing — over a weekend. The part with no screen: an MCP server that runs the books when I just talk to Claude. Cleaned up and open-sourced, MIT.
Read I Vibe-Coded an Open-Source Bookkeeping App with a Claude MCP Server on Medium · Code on GitHub
So I built an open-source Slack bot that watches spend, explains anomalies in plain language, and learns my normal — no per-seat fee, and it gets quieter the more you correct it.
Read Democratizing Cloud Costs: An Open-Source, AI-Powered AWS Cost Monitor for Slack on Medium · Code on GitHub
We learned that running a production RAG chatbot. The behavioral-testing approach we built, and why anyone shipping LLM features needs one.
Read AI Chatbot Behavioral Testing: Generative AI and RAG Applications on Medium
Like financial debt, the question isn't whether to take it on — it's whether the return beats the interest. When to incur it, when to refinance, when to pay it down.
Read Strategic Technical Debt: A CTO's Guide to Smarter Development Decisions on Medium
The most expensive call a tech leader makes with the least to go on. I've run it twice — once it built a thriving company, once it delivered and missed the business return. A spreadsheet model for the decision, where the value isn't the model's accuracy, it's the discipline of building it.
Read A White-Knuckle Decision: Refactor or Rewrite on LinkedIn · Spreadsheet model on GitHub
Some need one for a season, some for a single hard problem, some just often enough to avoid expensive mistakes. When the fractional model fits — and when it doesn't.
Read How You Know You Need a Fractional CTO on Medium