street-agent is a FERROUS Labs project run by Dan Barron. This page explains who I am, what FERROUS is, and what the project is for.
I'm a licensed Professional Engineer in New York and the principal of Barron Engineering, a civil/structural/geotechnical practice that does NYSDOT permit work, site plan reviews, and structural design for residential and commercial clients. That's my day job.
In parallel, I build software. For the past two years almost all of that software has been agent infrastructure in Rust — tool registries, capability systems, risk engines, reasoning loops. The civil engineering discipline and the agent engineering discipline turn out to be more similar than anyone expects: both care deeply about load paths, failure modes, and what happens when the probabilistic part of the system meets the unforgiving part.
street-agent is where the two halves of my work intersect — a real engineering project running against a real benchmark, with a design vocabulary that belongs more to structural engineering than to fintech.
FERROUS is my Rust AI agent company. The practice covers two things: building production-grade Rust agent infrastructure for clients, and publishing the patterns and frameworks that make that infrastructure work.
The name is deliberate. Rust is the language, iron is the metaphor, and the branding is industrial rather than sleek — because the thing I care about building is structurally sound, load-tested, and boring to operate. Flashy agents that impress on demos fail in production. FERROUS is the practice of making the opposite.
street-agent is a FERROUS project. Its public dashboard and engineering content are the top of FERROUS's funnel. If you're a prop shop, a fund, a fintech, or an engineering leader who needs this kind of infrastructure built for your own use, FERROUS is who you hire to do it.
Anyone can claim Rust agent expertise. Few will run one against live capital and publish every trace. This project is the public demonstration of what FERROUS can build.
Open questions about LLM reasoning under structured constraints — what breaks first, which rejections dominate, whether self-grading produces real learning. Live operation answers questions that paper-trading cannot.
A live agent produces a steady stream of genuine engineering content: architecture posts, postmortems, observations. That content is the actual marketing for FERROUS, and the dashboard is what makes people read it.
For FERROUS work — Rust agent infrastructure, risk engines, tool registries, MCP servers, or custom development — email works best.
For questions about street-agent specifically, public Q&A happens on the engineering blog. I don't offer private trading conversations or one-on-one advice.
For press, research collaboration, or speaking, same email.
Thank you for reading. Now go watch the agent work.