AI Builders · May 2026
The episode was about who ends up maintaining software that non-engineers generate, and McGirr reached for a definition of production that has nothing to do with uptime or polish. He backed it with an old rule from Cox Automotive: ship something in your personal namespace and you wear the pager for it yourself. He went on to say that vibe coding gets teams to that wall faster and with more force.
Transcript
files to some developers and said, can you make that? And the developers are like, no.
It's the same. Yeah. And
I think it's, you know, what is production, actually? And I've always thought of production as a function of how many other people than the creator can fix and iterate the product. And the ideal state of that is a large number. Like, you know, if it's a risky bit of software creating a lot of value, we want a large number of people other than its original creator to be able to fix it and to be able to run it. So remember, we had that rule at Cox Automotive that if you ship something to someone in your personal namespace, so dev underscore Sean, you wear the pager for that, right? And if it's a board member or whoever, none of the rest of us are obligated to support that. We just pass it on to you. If you convince your local team, dev underscore DS dev underscore BI, if you document it well enough, if they understand it well enough and you convince them, then they will support it. But prod, you have to convince data engineering to support now.
Yeah.
Right? And that's quite a jump. And I think vibe coding just gets us to that brick wall faster and with more force. Part of the force behind us is the potential value of if we can solve that. I think on the front-end side of things, like I think that's a very solvable problem. Like I've had people show me, oh, I put our design system in our core. Artifacts into Claude or Copilot or anything else. And now every front-end that people Vibe code in there is just already conforms to all of our standards and our best practices.
Yeah. And