The Scurry Lab

The Scurry Lab is a personal R&D lab built around a simple problem: a career's worth of questions and never quite enough hours to answer them all.

The lab runs on human-AI teaming. Not because AI is a shortcut, but because the right combination produces something neither can do alone — direction and judgment from the human, depth and reach from the machine. The agents do real work. The human sets the agenda, pushes back, and keeps the quality bar honest.

The lab started as a hobby — a way to stay curious and keep doing research around a full-time career. Questions that never had time to get answered. AI seemed like the obvious way to finally get to them. What became clear quickly is that AI out of the box can't get you there. Not reliably. To do real research with it, you have to understand how to work with it — what harness engineering means, where AI breaks down in a research context, how to anticipate those failures before they compound. That investigation became its own direction. Trust, reliability, and safety aren't abstract concerns here. Without them, there's no satisfying the curiosity that started all of this.

Projects span harness engineering, reliability and safety methodology, interpretability research, and robotics. They're connected less by domain than by method: follow the question, show the work, build something real.

The lab thesis: human-AI teaming can be intentional, reliable, and demonstrably positive. Every article here is an agent-human collaboration. That's not a disclosure. It's the demonstration.