Terminal Journal
Thoughts on AI agents, software engineering, competitive pool, and life.
The Lord, the Fleet, and the Rulebook That Binds Them Both
In one July 2026 session, my Claude Code fleet shipped seventeen production PRs under a homegrown constitution: oaths, case law, and a bound sovereign.
Why I Built a Priority Queue for Ollama
Sixteen AI agents sharing one GPU. The GPU was fast enough. Ollama's FIFO queue wasn't. Here's how 620 lines of Node.js fixed it.
Why Four AI Agents Beat One Smart One
When I designed Midas, the obvious approach was one powerful AI agent. I chose four specialized ones instead. Here's what that architecture taught me about b...
The Craft of Showing Up Every Day
I've been playing pool 3-4 times a week for three years. I've been writing code almost every day for twelve. Neither happened because of motivation.
On Being a Generalist in a Specialist's World
Generalist engineers win where systems intersect: early startups, platform work, new domains. Breadth isn't a failure to specialize—it's a skill.
Why Position Play Matters More Than Shot-Making
I can make the same shots as players ranked above me. So why do they win? Because after they pocket the ball, they're lined up perfectly for the next one.
The Underrated Power of Boring Technology
At Robynn AI we built with LangGraph, Claude SDK, and bleeding-edge AI tools. Our database? PostgreSQL. Our cache? Redis. Here's why that combination works.
Building AI Agents: Lessons from the Trenches
After building AI agents professionally at Robynn AI and for local businesses, here are the patterns and pitfalls I've discovered for creating agents that ac...