Early prototypes teach faster than strategy decks. These notes cover what we learned before we had a clean story — including two tools we shut down.
Prototype A: the dazzling tutor
It answered beautifully. Testers loved the first session. By the third, they were copying outputs into other apps and skipping the understanding step. Capability retention collapsed. We retired it because it was winning demos and losing the week after.
Prototype B: the silent coach
It almost never spoke first. Testers called it “broken” in week one and “the only one I’d keep” by week four. The problem was onboarding: silence without a legible contract feels like failure. We kept the temperament and rewrote the first ten minutes.
Why spectacle wins early
Spectacle is measurable in the room. Usefulness is measurable in someone’s Tuesday. Our testing culture had to change before our products did: longer protocols, tool-removal checks, and a rule that a standing ovation in a demo is not a shipping criterion.
We still catch ourselves optimizing for the wrong applause. Publishing the retired work is one of the ways we correct.