blog
Intelligence Is Becoming Abundant
Open weights are making useful intelligence cheap enough to become ordinary infrastructure.
I fine-tuned two small medical LLMs
I fine-tuned two small medical LLMs to reason, then tested two contradictory 2026 calibration claims. The answer depended entirely on how I asked the model how sure it was. All on a free GPU.
How to secure skills?
How I started thinking about agent skills as untrusted code, and the four things that would make one safe to run.
You Saw It Coming
On arriving at ideas independently, prediction as a skill, and why seeing it coming means nothing if you don't act.
What is elegance?
One word changed how I talk to AI. It took me a while to figure out why.
Why Taste Matters
What taste actually is, why it's the thing that separates good work from great work, and how to develop it.