The Stack Doesn't Matter Anymore
Most teams I see using AI are shipping faster than ever. They’re also accumulating technical debt faster than ever.
Those two things are connected.
The first instinct is always the stack. Wrong framework. Wrong language. Should’ve picked TypeScript from day one. But that debate was always a distraction. The pace was just slow enough to hide it.
AI took the cover away. Accelerate a team with weak thinking infrastructure and you don’t get better output. You get the same bad decisions, faster.
The way I see it, good thinking infrastructure comes down to three things.
Contracts. Explicit agreements about how your system behaves. What goes in, what comes out, what it never touches. AI writing code against a real contract has something to work with. Without one it’s pattern-matching and hoping.
Decision frameworks. How your team actually makes calls. Not every decision needs a meeting. It needs a principle written down somewhere. AI can execute against a principle. It can’t figure out your values for you.
Thinking systems. How work flows, how problems get broken down, how context moves through the team. If you can’t explain it, AI can’t extend it without breaking something.
AI is not an architect. It’s an accelerant. Better models help. Better tools matter. But a faster engine pointed the wrong direction just gets you more lost.
The thing AI can’t give you is judgment. That comes from building things, watching them fall apart, and understanding why. The kind of experience that tells you what’s actually worth building in the first place.
You bring that. AI amplifies it. Not the other way around.
Before you touch your stack, audit your thinking. Do you have contracts that make intent explicit? Decision frameworks your team actually uses? Thinking systems clear enough that AI can extend them without breaking them?
Delegation only works when the infrastructure exists for a system to behave correctly. A team, a codebase, an AI agent. Same principle every time.
Without that foundation, AI just speeds up the mess.
Build the infrastructure first. Then let it run.
- Iverson