AI coding has 10x my ship velocity. I’ve crossed a threshold where the bottleneck isn’t writing code, it’s knowing what code to write.
This changes my job to be a reviewer and architect, for the most part.
The new skill: iterative interrogation until you reach the simplest working solution.
Here’s what I’m seeing in practice
An agent’s first solution typically optimizes for completeness. It builds the general case. New tables, abstraction layers, API endpoints—all “technically correct.”
But correctness isn’t the goal. The right solution is.
The pattern that works
- Round 1: Agent delivers a working solution
- Round 2+: You probe the assumptions. “Why this architecture?” “What constraint drives this complexity?” “Can we solve this at a different layer?”
- Round n: Agent simplifies. “You’re absolutely right.”
When the agent acknowledges a simpler path—that’s your signal. Not because it agreed with you, but because you’ve pushed past default complexity to find the essential solution.
A recent example for building streaks
- Initial solution: 3 database tables, multiple API endpoints, client-side state orchestration
- Final solution: 3 UI components, 1 page route, existing data model
Same outcome. 80% less code. Zero new infrastructure.
The difference wasn’t the agent’s capability. It was knowing which questions to ask.
What’s changed
We’ve amplified solution exploration by orders of magnitude.
What hasn’t changed
Someone still needs to judge what “done” looks like. That’s the irreducible part.
The new model
Generate options → Interrogate → Simplify → Ship
The discipline now is subtraction, not addition.
If your codebase is growing faster with AI than it did without it, you’re not asking enough questions. The agents can explore the entire solution space. Your job is knowing when you’ve found the right corner of it.
Building with AI isn’t about accepting what works. It’s about pushing until you find what’s right. The “You’re absolutely right” moment is where velocity meets judgment.