← Work

pi-bot

Phase A - in build

An embodied, local-first AI agent driving a rover, built around a safety layer that can refuse unsafe commands.

A rover that runs its own AI brain on local hardware and moves under its own decisions. The part I care most about is the layer underneath it: a capability-based safety system that gates what the agent is allowed to do, and refuses anything outside those limits.

I’m building it in the open and keeping the log here as it comes together - the wiring, the dead ends, the decisions that turned out to matter.