Travel

Wimbl: A Recursive Family of Agents for Football Travel Wimbl is an event-led travel discovery system built around a family of specialized AI agents. It helps travelers turn an interest in attending a football match into an evidence-backed trip plan. Its hierarchy has three levels: a Sports parent that maintains general travel preferences and verification standards; a Football parent that contributes football-specific knowledge about match-day timing, ticket-sale patterns, and trusted sources; and short-lived trip agents that specialize in tickets, flights, hotels, local transit, price monitoring, and coordination. Wimbl combines web discovery, feasibility checks, scoped memory, and evaluation of candidate improvements. Search results are treated as evidence rather than guaranteed availability, missing prices are not invented, and lessons are evaluated before being considered for broader reuse. This creates a path toward agents that learn from experience while preserving provenance, scope, and human oversight. The project uses a FastAPI application, You.com for web discovery, and Crusoe Managed Inference with NVIDIA Nemotron models. The current implementation includes the agent hierarchy, memory and evaluation scaffolding, and a working API surface. The longer-term vision is a recursively improving family of travel agents whose specialized children can contribute validated lessons back to their parent agents.