The argument / 01
The constraint
moved.
A practice's value depends on the constraint it solves. Follow the change from engineering effort to verification, judgment, and trust.
01 / Rationing was rational
Building less wasn't just cheaper. It was how you learned.
Piece 01Lean made expensive engineering answer to evidence. For a team that could afford only a little building, a smaller experiment was a sensible discipline. The original logic deserves to be understood before it is replaced.
02 / The cost moved
Building is the part that got cheap. Being believed didn't.
Piece 02When AI changes the cost of building, a full product becomes possible before it has earned anyone's trust. The scarce work moves to proving capability, deciding what deserves to exist, and controlling who encounters it. Build the product you believe in, then widen exposure against written criteria.
03 / The verdict stays outside
You can't spin up a customer.
Piece 15Internal checks can establish that a product works. They cannot establish that anyone wants it. Customer behavior remains evidence you have to earn outside the building. A dark launch needs an exit condition; darkness alone is no achievement.
A failure worth keeping / Piece 03
The human caught white text sitting on a white background.
In piece 03, Daniel describes ad renders that passed the mechanical checks. A human found the text unreadable. Legibility had never been part of those checks. The catch became a mechanical contrast guard.
- Mechanical checksPassed the rules they had.
- Human useFound what the rules missed.
- New checkKept the lesson in the work.
The fix retires an incident; the check retires a category.From piece 03
Two loops. Different verdicts.
Prove that it works.
Then learn if it matters.
Inside the build
- Plan
- Build
- Verify
- Compound ↶
The market loop
Customer evidence
A separate verdict.
Never an automatic pass.
The inner loop tells you it's real. Only the market tells you it matters.Piece 04
The limits belong
in the argument.
The argument has a scope
If engineering remains your scarcest input, rationing it can still be rational. This argument addresses teams where AI has changed that constraint. Maintenance, verification, and customer trust still cost something.
Full ambition still needs judgment
Cheap building does not make every idea worth building. Customer development, positioning, pricing, and jobs to be done still require evidence from outside the building.
Checks move into the work
Moving stop authority to where a bad action happens is the principle. Turning off paid CI is a contingent example from one operation, not a universal prescription.