Work
Things I've built
Open-source projects and experiments, each with the reason it exists.
- 01Rust · Desktop
Cortex
A local-first brain for you and your agents. It remembers what happens on your screen, lets you ask about it later, and spots reminders and tasks in your conversations.
Why I built it
Memory tools store facts but don't help agents split up and pass work between each other. Cortex is a shared task and memory space, so a task started in Codex can move to Claude with its context intact.
- 02Python · LangGraph
Agent Loop Recovery
A reproducible LangGraph experiment that detects a stalled support agent and recovers through bounded state repair.
Why I built it
Agents get stuck retrying the same failing call, such as querying an obsolete shipping label. I wanted a measurable demo of detecting no-progress loops and repairing state safely instead of burning the budget.
- 03Swift · iPad
OpenDisplay Touch
A personal fork of OpenDisplay that adds an iPad system keyboard, three-finger Spaces gestures, and full-screen window controls.
Why I built it
Using a spare iPad as a second Mac monitor is great until you need to type or switch Spaces. This fork makes the iPad usable as a touch-first extension of the desktop.
- 04Python · Evals
Agent Recall Lab
Evidence-backed coding-agent memory: trace ingestion, reviewed lessons, recall before actions, and reproducible evaluations.
Why I built it
To test one question: can a coding agent avoid repeating a mistake if a lesson from a previous session is retrieved before it runs a command?