Dyadem
Anonymous community feedback platform where AI learns what to ask. Built for the Gemini 3 Hackathon. Residents answer three quick questions about cost-of-living pressures, then Gemini generates personalised follow-up questions based on both their answers and patterns across the entire dataset. The result is a living "Community Voice" narrative that evolves as more people contribute.
Challenge
Traditional surveys ask everyone the same questions regardless of context. Focus groups capture nuance but don't scale. Communities need a way to share experiences that adapts to what's being said — and synthesises many voices into a collective insight.
Solution
Built an adaptive feedback loop: fixed questions establish a baseline, Gemini generates tailored follow-ups using structured output, a theme extraction layer identifies patterns across all submissions, and a separate high-reasoning Gemini call produces a narrative insight grounded in real data. The questionnaire gets smarter as the dataset grows — questions at submission 200 are fundamentally different from submission 5.
