Stripe has started to show what the next engineering organization looks like.
More than 1,000 pull requests per week. Written by agents. In production. In a codebase with hundreds of millions of lines, where mistakes are expensive.
The line that matters is simple:
Developer attention is the scarcest resource.
That is the actual shift.
The Bottleneck Moved
For a long time, most organizations behaved as if code production were the scarce thing.
Need more output? Hire more developers. Need more throughput? Add more teams. Need more certainty? Add more routing, more coordination, more people translating between business and implementation.
AI changes that equation.
Code is cheaper. Translation is cheaper. Iteration is cheaper.
But attention does not scale the same way.
Judgment still sits with humans. Product understanding still sits with humans. The ability to decide which problem matters, which tradeoff is acceptable, and which output is safe enough to ship still sits with humans.
That means the new organizational question is not “how do we get more code written?”
It is:
Where are we still burning scarce human attention on work that a system could already absorb?
What Stripe’s Minions Make Visible
What makes Stripe’s Minions work interesting is not that AI writes code.
That part is table stakes now.
What matters is the operating model around it:
- flaky tests get triaged automatically
- the next likely action is already proposed
- human attention is reserved for approval, review, product sense, and other judgment-heavy steps
That is not just productivity tooling. That is attention allocation.
The system takes over deterministic routing and repetitive handling. Humans stay focused on decisions that actually require context and responsibility.
I See the Same Shift in Practice
I have been learning the same lesson very directly over the past months.
I delegate more than 99% of my coding work now.
That does not mean I disappeared from the process. It means my role moved.
I plan. I observe. I review. I steer. I decide.
The leverage is brutal because the expensive part is no longer typing or implementation mechanics. The expensive part is attention.
The people who can hold system context, make pragmatic tradeoffs, and decide what matters become more leveraged. The people whose main function was translation, routing, or manual throughput lose relative importance fast.
The New Question for Teams
The old question was:
Do we need more developers?
The new one is:
Where are we wasting human attention on routing, translation, and handwork that a system should already be taking over?
That is what teams need to audit now.
Not just tool adoption. Attention flow.
Because code scales. Attention does not.