Most of what I write here comes in series — each one builds an argument across several posts, and they read best in order. Start anywhere below.

Building an AI Agent for My Family

A build log. Seven posts on making a bilingual family assistant with local, open-source models — the design decisions that mattered more than the model.

1. I Built an AI Agent for My Family
2. A Voice Nobody Wants to Mute
3. Propose, Don’t Act
4. Find, Surface, Hand Off
5. A Bot That Teaches
6. The Homelab Behind the Family AI
7. Homework Stays Private

Agentic AI and the Enterprise Stack

Where agents are reshaping enterprise software, and what it does to platforms, pricing and the people who sell advice.

1. Agent Swarms Are Eating the Stack Top-Down
2. The Substrate War
3. Buy, Build, or Orchestrate
4. Service-as-Software
5. The Frontier Labs Are in the Squeezed Middle
6. Who Advises the Disrupted?
7. Agentic AI Is Not Just LLMs and Tokens
8. A Product Playbook for Agentic AI
9. Why Agentic AI Transformation Is Hard
10. Give Your Coding Agents a Process
11. A Longer Leash
12. Entity Resolution Is the Bottleneck

Cheap to Be Wrong

A six-part architecture series, plus a companion: where to put the model, where to put code, and why the line between them is the whole design.

1. Put the Model Where Being Wrong Is Cheap
2. Use the Model to Write the Parser
3. Where the Model Is Genuinely Load-Bearing
4. An Agent Is a State Machine, Not a Loop
5. The Model Upgrade That Quietly Breaks You
6. The Model Is the Least Defensible Part of Your Agent
7. Does a Smarter Model Change Any of This?
8. What a Good Eval Set Actually Looks Like

I’ve Seen This Wave Before

Lessons from founding a startup in the Big Data and Cloud wave, applied to the AI wave — and why the sequence rhymes.

1. I’ve Seen This Wave Before
2. Start With the Problem, Not the Wave
3. Being Early Is Not an Advantage
4. Services Before Product-Market Fit
5. When Your Partner Ships Your Product
6. What Do You Charge for a Data or AI Platform?

When Your Product Becomes a Tool

What happens to incumbent software once its capabilities are callable — the mechanism, the diagnostic, the playbook, and the buyer’s side.

1. When Your Product Becomes a Tool
2. What Dissolves When You Become Callable
3. Substrate or Façade
4. The Incumbent’s Playbook for an MCP World
5. Buying in an MCP World
6. What Agents Cost to Run
7. Your MCP Server Is an Attack Surface