Last week I had a surprising realization while working with AI agents.
I stopped caring so much about the structure of the code.
Not because structure doesn’t matter—but because I’ve started to think differently about how we build software in an AI-assisted world.
The Old Way
I’ve spent decades building and scaling systems with clean architecture, clear boundaries, and carefully crafted abstractions. That stuff still matters.
But now, when an AI agent sends me a pull request, and it’s structured in a way it understands—even if it’s not quite how I’d do it—I no longer feel the need to correct it.
Why?
Because I can see where this is going.
The New Mindset
AI agents are not just tools. They’re evolving participants in our workflows.
If I can provide structure they can reason about—and if they can return work in a format I can understand and review—then we’re not just writing code anymore.
We’re building shared understanding.
I still review every line of code. I make sure it works. I ensure it aligns with business goals and product intent.
But I’ve shifted focus.
What Really Matters
It’s not about perfect code.
It’s about:
- ✅ Clarity of purpose
- ✅ Testable boundaries
- ✅ Reviewable PRs
- ✅ Reliable outcomes
- ✅ Context that flows between roles
The code is one part of the system. But the design system, the spec process, and the ability to have contextual conversation about what's happening—that’s where the real value is.
The Future is Conversational
As this work progresses, the way I communicate with AI is starting to feel a lot like the way I’ve mentored junior engineers.
Explain why.
Ask questions.
Review together.
Agree on intent.
Then ship.
It’s not about controlling every line.
It’s about building a working relationship—with tools that are starting to feel a little more like teammates.
And honestly?
That’s exciting.