AI travel planning
Odyssey
Planning a trip is mostly research: which neighbourhood, which days, what is near what, how much can fit before everyone is tired. Odyssey compresses that work into a sentence. Describe the trip you want, a long weekend in a city with good food and one big hike, say, and the planner returns a structured, day-by-day itinerary in seconds.
The part we cared most about is how the plan arrives. Rather than a spinner and a wall of text, the itinerary streams in as it is generated, day by day, so the person watching sees the shape of the trip form and can start reading before it is finished. It makes a wait of several seconds feel like a conversation rather than a delay, and it was the detail that took the most engineering to get right.
A first draft is rarely the final one, so the plan is a starting point rather than an answer. A traveller can keep refining it, swap a day, push a museum to the morning, ask for something quieter, until the trip looks the way they had imagined it. Each pass keeps what was working and changes only what was asked.
Odyssey is a full web application with user accounts, so plans are saved and can be returned to. The entry screen is deliberately spare: the logo, the line “Plan your next adventure” and a sign-in over a mountain landscape. The interface stays out of the way of the itinerary, which is the thing people came for.
It is the kind of product we build when a client has a product idea of their own: an interface designed around a single core interaction, an AI model behind it, and the streaming, accounts and infrastructure to make it feel instant.