These directions do not follow the current AI design scheme. They use behavioral economics as wayfinding: one default path, low choice competition, clear progress, and support where uncertainty occurs. Every direction starts with the same first move: define the problem before looking for obstacles.
A measurement instrument. WIMD starts by getting the problem sentence stable enough to measure against reality.
Do not solve it yet. State the problem you think you are solving.
Palette: white, black, blaze orange. No softness.
Messaging: define, measure, test.
Risk: can become too severe if every sentence sounds like an audit.
An inspection surface. It holds the problem still before asking what boundary is shaping it.
Hold the named problem in place before inspecting the constraint.
Palette: off-black, white, hazard yellow.
Messaging: state, inspect, reduce.
Risk: intensity must not become punitive.
A field of indirect movement. It starts by naming the problem, then uses small tests to let reality correct the angle.
The first angle is the problem sentence. The test comes later.
Palette: deep green, near-white, signal orange.
Messaging: name the problem, then move at the angle that produces evidence.
Risk: precession must remain plain enough for first-time users.