Short letters on how AI agents actually work when money is on the line.
Walkthroughs of systems I've shipped for clients. What worked, what broke, and the part I redesigned three times before it held.
How to structure prompts, route between models, handle failure gracefully, and keep agents from spiraling into expensive nonsense.
Make, n8n, GoHighLevel, Claude, OpenAI. When to reach for which. The tradeoffs nobody mentions in the tutorials.
Half-formed ideas. Things I'm stuck on. This letter doubles as my thinking-out-loud space, and you get a front-row seat.
A walkthrough of the three places I usually find the bottleneck when a client's automation slows to a crawl.
Model selection is the single biggest lever nobody talks about when building reliable agents for real client work.
A case study on the first real automation I built. What the problem looked like, what I did, and what I would do differently today.