Values
Experiments before predictions
Take a position, build it and find out.
Rework is cheap now, so treat an initial design as a set of hypotheses to test. Give each one a statement of how you’d know it’s wrong. If you can’t picture the answer yet, ask the agent for a few options and converge on the one that works.
The trap
You analyze and debate until you’re sure, as if change were still expensive. Or you do the opposite and wait for a design to emerge without taking a position.
When prediction is right
- Feedback is expensive and you already know the answer. Industrial measurement equipment needs a careful specification in a way an online store does not.
Where it shows in the grid
“Humans spend their time on problems and experiments, not on reviewing output.”