Writing
Notes, devlogs, and side quests
What I'm building, what I'm learning, life, and everything in between.
OpenRadView: an app I hope you never need
A free, open-source DICOM viewer for seeing your own CT, MRI and X-ray images on your computer. A personal research project, not certified medical software.
tokenstat: from a token heatmap to a home base for coding agents
tokenstat started as a CLI that drew my AI usage as a heatmap. Three months later it runs coding agents in terminals and chat on a Mac or server, hands work between agents, and reaches them from an iPhone, iPad or Android phone over an end-to-end encrypted relay.
A Claude Code agent fleet: 500 pages in 15 languages
I started a Claude Code agent fleet from my phone, went to sleep, and woke up to 500 SSR pages in 15 languages. A devlog on running an overnight fleet of Opus, Sonnet, and Haiku agents.
The Claude Code habits quietly running up your usage
Most of what makes Claude Code slow and expensive is habit, not the tool. Here are the seven that cost me the most, each with a one-line fix and where to dig deeper.
The tools I use for agentic work
Agentic work has quietly become the new IDE. A tour of what I actually reach for: the coding agent itself, the terminal I run it in, the editor I read the diff in, remote access, local models, and the app that keeps the Mac awake through it all.
Running Claude Code on local models with LM Studio and Ollama
The Claude CLI is just a client for the Anthropic API, so anything that speaks that API can sit behind it. Point it at LM Studio or Ollama and you get the same terminal workflow on a local model. The same trick runs local models, voice, and memory in MyAgens.