Limen: I Put a Camera at My Window and Asked Who Owns What It Sees
This week is Fix, Hack, Learn at Microsoft, and my project is a local-first street-vision system I am calling Limen, the Latin word for threshold. An Arduino board watches my front window, a vision model on my own MacBook Pro M3 Max names what crosses (person, vehicle, wildlife), and a human confirms the label to build a training set I actually own. No cloud, no subscription, no frame leaving the house. Today I ran a scripted 90-second proof of concept: I walked up, drove in, parked, and walked to the door, and every one of 180 frames was auto-labeled locally in about 8.4 seconds each. Here is what worked, what did not (plates are still unreadable), and the one hard constraint that shaped the whole architecture: my evals rig is sacred and this project was never allowed to touch it.
This week is Fix, Hack, Learn (FHL) at work (Microsoft), our internal week to chase a build we care about. Last FHL I worked on my agents and agentic loops, helping me scale myself and my work as a product manager. This time my project is personal, and it starts with a question I could not stop turning over: when a camera watches the front of my house, who owns what it sees?
