The highest-leverage act in building AI systems is understanding the mind that uses them.
Ideas on cognitive architecture, agentic workflows,
and why specification is the new literacy.
Essays and field notes from building AI infrastructure — on what worked, what I learned about my own assumptions, and what the evidence made me change.
Read the first piece Latest: Be the advocate for your agent’s UXField notes
What happened when the ideas met the work
-
Be the advocate for your agent’s UX
My first week of Mega.dev, a four-week online program for developers who already build with AI, strengthened some of my ideas and reminded me of the importance of having external eyes look at how my system works.
-
The Bug Is in the Report
Checking what agents claim about their own code
In one run of coding agents, 8 of 26 serious bugs contradicted a sentence the agent had written about its own code. What I changed in the agents' instructions afterward, and what it did and didn't fix.
Ideas
In suggested reading order
How Your Mind Actually Works
A framework for discovering cognitive operating characteristics and designing systems that fit
Most design asks "what do you want?" That's the wrong question. The right question is: how does your mind actually move?
Why AI Workflows Click for Some People
It's not about intelligence. It's about cognitive fit.
The people who take to AI workflows quickly have — usually by accident — found one that matches how they think.
7 min readThe Expertise Paradox
Why knowing how to build can make it harder to let go
The people thriving with AI tools aren't always the strongest engineers. That's not an accident — it's a pattern worth understanding.
8 min readThe Four Questions
A framework for understanding any AI agent system
Every agent system reduces to four questions. The answers show where it's likely to work and where it's likely to fail.
8 min readSpecification Is the New Literacy
When a first implementation takes minutes, the bottleneck is knowing what to build
The shift from "can I build this?" to "can I specify this precisely enough that it gets built correctly?" changes everything.
9 min readWhen Systems Fix Themselves
Building an auto-healing architecture by composing the patterns that couldn't do it alone
Five campaigns, five phases, and a loop designed to close. Building a self-correcting dispatch system, what it revealed about composing agentic patterns, and where it fell short.
9 min readDispatches from the Build
What I learned from building AI infrastructure sixteen hours a day for eight weeks
The velocity was real. So was the risk of burning the candle at both ends. Here's what sustained it — and what almost didn't.
10 min readAbout
I've been building software for twenty-five years across hobby retail, video games, and maritime shipping. Same instinct every time: find what's broken, make it work for people. In late 2025 I started building AI infrastructure. These essays came out of that work: what it looks like when you take specification seriously, and why it starts with understanding how you think.