# I build with agents. The hard part is knowing what to ask.

> Ben Schmidt is a software architect and founder (HeyLoopy; EIR at CMU's Corporate Startup Lab) who builds with AI agents daily and writes about the scarce skill of the moment: knowing what to ask for, and knowing when the confident answer is wrong.

Source: https://iambenschmidt.com/about/

<script type="application/ld+json">
    {
      "@context": "https://schema.org",
      "@type": "Person",
      "name": "{{ .Site.Params.Author.name | safeJS }}",
      "image": "https://iambenschmidt.com/ben/ben_010.png",
      "url": "https://iambenschmidt.com",
      "jobTitle": "Founder & CEO",
      "alumniOf": "University of Pittsburgh",
      "worksFor": {
        "@type": "Organization",
        "name": "HeyLoopy"
       }
    }
</script>

I build software with AI agents every day, and I write about the part the demos skip. The tools will build almost anything you describe, so the hard part is no longer building. It's knowing what to ask for, and knowing when the confident answer is wrong.

A couple of teams I advise recently asked me the same thing: how their developers should ship code, the path from someone's laptop to the live product. It's a question you can hand an AI and get a hundred confident, contradictory answers to, because it's almost all trade-offs, not correctness. What settles it isn't "is this right," it's "what are we optimizing for, and who gets to approve what goes live."

That's the pattern under most of it. "Build me the thing" is an engineering request full of trade-offs the agent will resolve silently if you let it. The people who do this well haven't memorized every framework. They know enough of the shape of a problem to ask the right question and smell a wrong answer. Most of what I write is aimed at handing you more of that.

I have been building for a while. I co-founded RoadBotics, an AI company for infrastructure, ran it as CEO, and sold it to Michelin in 2022. Now I build HeyLoopy, software that keeps everyone on a team, people and agents alike, working from the same up-to-date facts instead of stale docs and guesswork, and I advise engineering teams making the jump to AI-native development, the way I work myself.

Further back, I was a recovering brain scientist. I did a PhD on how networks in the brain communicate, which is where I first got my hands on machine learning, years before every company claimed to. It left me sure of one thing: the real answers are almost always less magical than the headlines promise.

I have been sold plenty of that, which is why I don't sell any. Mostly I write to hand you the questions that catch a confident wrong answer before it ships. That's what **[The Shape of It](/essays/building-in-the-open/)** is: short, deep guides that teach one fundamental of how software actually works, then give you a tool to point your own agent at, whether or not you code.

I also keep a research imprint, **[Oakmont Lab](/oakmontlab/)**. One bench checks published AI-in-learning claims against their own data; another is machine vision and sensor fusion for real-world robots. A hobby with standards.

That robotics thread keeps pulling me back: computer vision at RoadBotics, a robotics board seat now, sensors and planning in the lab. It's where more of my attention is heading.

## The short version

- **HeyLoopy** · Founder & CEO. Software that keeps a team, and the agents it uses, working from the same current facts.
- **CMU <a href="https://www.cmu.edu/swartz-center-for-entrepreneurship/mentorship-and-programs/csl/index.html" target="_blank" rel="noopener">Corporate Startup Lab</a>** (Tepper) · Entrepreneur in Residence, working with fellows on intrapreneurship.
- **<a href="https://sillysurfacing.com/" target="_blank" rel="noopener">Silly Surfacing</a>** · Board member. Early-stage robotics.
- **RoadBotics** · Co-Founder & CEO. Computer vision for infrastructure. WEF Technology Pioneer. **Acquired by Michelin in 2022.**
- **Michelin Mobility Intelligence** · CTO. Led the integration across US and France teams.
- **PhD, Bioengineering** (University of Pittsburgh) · studied how brain networks communicate, with machine learning on neural activity.

## Before you go

This is where I think out loud. The best way to keep up is right here, and I post to LinkedIn now and then: <a href="https://www.linkedin.com/in/iambenschmidt/?utm_campaign=iabs&utm_medium=web&utm_source=about" target="_blank" rel="noopener">follow along</a>.

If you're in Pittsburgh and working with agents, building with them or putting them to work, I keep a small table of people doing the same. Come say hello.

And if keeping your team and its agents working from the same facts is the problem you're chewing on, that's what I build at <a href="https://heyloopy.com?utm_campaign=iabs&utm_medium=web&utm_source=about" target="_blank" rel="noopener">HeyLoopy</a>.


---
This is the plain-markdown twin of https://iambenschmidt.com/about/. The whole site offers one: append index.md to any page URL. Index for agents: https://iambenschmidt.com/llms.txt