I wrote this in November. I had no proof. Three weeks later I learned I was right.
The skills that made you a good manager — architectural judgment, systems thinking, process building, knowing what good looks like — are exactly what AI-first development needs.
The prediction underneath it was specific. The best people for this aren't engineers — they're former managers: product managers, engineering managers, people who stopped writing code and spent years learning how to define what done looks like.
I believed it because I had watched it in my own hands. August 2024, at my last company, before any of this: my team quoted seven days for a classifier that would sort our salespeople's inbound email. I owned product engineering and I owned IT, so I knew most of that seven days was getting access, not writing code. I sat down with Claude in a browser tab, before Claude Code existed, and had it working in seven hours. Then I ran the same work through the team's estimation process — with the thing already built — and they estimated it at seven days, all over again.
I am not a good developer. I hadn't shipped production code in fifteen years. Go look at my GitHub: one dot. What I know how to do is solve problems: break them down and decide what matters first. That is why I knew I could get this one done with the robot. Paired with a machine that could just build the software, I knew I was unstoppable. What I wanted after that was to make other people unstoppable. That is the job I am hiring for.
Back to November 2025. Three weeks after that post, the applicants showed up. I ran them through my own interview: build something with the robot, with me, in the room. Senior engineers, junior engineers, managers, same test. The engineers wanted to talk about the technology, how the robot would write it, what the code would look like. The managers went at the problem and would not touch the how until they could say what done meant. This wasn't a fluke. The pattern held. I wrote to Aaron that night: “The managers? They go ham.”
One of them was already at offer stage. His first pull request landed December 15. It added himself to the telemetry config so his output would show up on the board where we track everyone’s numbers. Nobody asked. He had worked out that if we could not see his work, we could not tell whether he was doing the right thing. So the first thing he built was the way to check him.
That is the instinct I was predicting. I didn't hire a coder. I hired judgment. The AI supplied the rest.
This isn't magic and I'm not smart. The robot is that good. Everyone is out hiring AI developers, developers who want to use AI. That's not what I'm looking for. I want the people who used to build and got pulled into management for their judgment, not because they wrote the best code. The bottleneck is knowing what to build, how to scope it, and when it's done. That's always been a manager's skill set.
I'm hiring three roles that assume this thesis is true. If you used to build things and you miss it — if you spent the last few years in standups about work you used to do — this is the opposite of that. We don't need you to manage a team. We need you to ship.
The bar is judgment and taste, proven by shipping code or by running teams. Title isn't the filter. Have your AI send me a synopsis of what you built in the past 30 days. If you haven't built anything, this job probably isn't for you.
Not a typical security role. Twenty-one people ship production code here, most of them through agents, and you own the controls that make that safe. You build them and you automate them, with the robots, rather than hiring a team to hold them up by hand. A control holds because it holds; the audit confirms what was already true.
Have your AI send what you built in the last 30 days.
Same attitude. Judgment and operational execution, plus the robots. New products on the horizon. The factory is v0.8. You would be building the rest of it.
Have your AI send what you built in the last 30 days. If you haven't built anything, that's an answer too.
You already know what to build. You want the bots so the people closest to the customer can operate at a different scale. You ship too.