Blog

Writing on building with language models. Ideas from building LLM features for real users, not from demos.

AI Isn't Making You Worse. Not Reviewing It Is.
The skill isn't rotting from AI use. It's rotting from skipping the check. How to make reviewing AI output cheap enough that you'll actually do it.
June 10, 2026
You Can't Review What You Can't Evaluate
Adding a human checkpoint to an AI workflow doesn't make the output trustworthy. The reviewer has to be able to evaluate what the model produced.
May 29, 2026
Your First AI Project Shouldn't Be an AI Project
The AI community has an on-ramp problem. The right starter project isn't an AI project at all.
May 16, 2026
Your Skills Folder Is a Junk Drawer
Most skills shouldn't exist. The energy the community is pouring into building and collecting them is mostly wasted.
May 13, 2026
LLMs are functions, not brains.
A case for treating LLMs as typed functions, not autonomous agents. With live demonstrations.
May 8, 2026
Agentic-First: Designing Systems for AI From Day One
Most teams hit a ceiling with AI and assume the model is the problem. It isn't. It's the system around it.
April 12, 2026