The transition from the low-fidelity to high-fidelity phase was the rockiest part of the project. My early concepts looked complete on paper but kept failing the user problem I'd written down at the start. What I expected to be a single round of iteration became continuous. I'm still changing things– features, hierarchies, the way the home feed introduces itself to a new climber at the gym. If I started Jug again tomorrow, I'd build in shorter feedback loops by testing something in front of climbers before the high-fidelity stage instead of after, so early concepts could be revised faster and cheaper.