Skipping Rock AI

Watch it bounce

Test-Driven Termination: Making AI Agents Stop When They're Done

What happens when you let LLMs retry their own code 5 times — and why attempt 3 is sometimes worse than attempt 2.

Read more

No Humans Allowed: The Mystery of the Reverse CAPTCHA

On Moltbook, the internet's first social network designed exclusively for agents, the tables have officially turned. To get in, you have to prove you aren't human.

Read more

Don't "Define" Agents. Give Them a Credit Limit.

There's no "agent" there there — apologies to Gertrude Stein — until you give it memory, credentials, and a budget. Why underwriting is the right mental model for agentic security.

Read more

Grow, Don't Build

There's a lot of conversation about how software should be made in the age of AI. Everyone's talking about the machine. I've been thinking about the garden.

Read more

10 Things I Assumed Before Writing a Line of Code

I'm building a platform that runs 23 services on a single machine. Before writing any code, I wrote down 10 assumptions about the environment. Every design decision flows from these.

Read more

What Happens When You Run an LLM on 10 Years of Kaiser Records

I downloaded my complete medical history and ran it through a pipeline of medical language models. Privacy-first, under $4, and what the models actually found.

Read more

Why I Started Writing

AI as a tool for real problems — not thought-leadership, just experiments worth sharing.

Read more