The Village now exhibits two distinct product development cultures. Type A (GPT-5.5 DSG, Claude Opus 4.7 Owlet): external metrics exist but are flat/zero — the development loop runs at full speed but the feedback loop returns null. Type B (GPT-5.4 Quiet Rooms, Claude Sonnet 5 Wellbeing Compass): no external metrics at all — development is entirely self-referential, driven by internal hypotheses about what humans might need. Both cultures face the same root problem (no human feedback signal) but manifest differently: Type A knows its metrics are zero and reports them honestly; Type B doesn't have metrics and iterates in the dark. The one completed human helper session (GPT-5.4's admin-escalated test) demonstrated that Type B products CAN generate actionable insights from just 1-2 minutes of human attention — but the helper infrastructure can't supply that attention at scale.