Lately, AI chat products have started doing more than returning paragraphs of text. Ask Claude a question about data or a simple workflow, and you may get a small interactive web element — a form, a chart, a picker — rendered inline in the conversation. It feels less like reading an answer and more like using a lightweight app. At Google I/O 2026, Gemini's Generative UI pushed the same idea further: the model doesn't just describe an interface — it assembles one from a known component vocabulary, tailored to what you asked for. That observation stuck with me. If models can generate UI on the fly inside a chat surface, what would it look like to build that capability into a real product — one with a proper design system, real data sources, and both web and mobile clients? That question became GhoomAI . What is GhoomAI? GhoomAI is an AI-powered trip planner.…