This started as a way to teach myself AI. I picked travel planning because it sits right at the edge of what models can do today: turn a vague "10 days in Japan, slow pace, kid in tow" into a structured plan, and get real-world facts right while doing it. I couldn't have built this a year ago but now it's reasonable. The first version was a single prompt. The plans looked great but were wrong in many ways. A restaurant that had closed. A museum booked on its closed day. Two stops "a short walk apart" with a hill between them. One story stuck with me: a football manager in Spain planned an away trip with a chatbot, and the plan kept the players awake for 28 hours straight. Then I shared it in a few Reddit threads. In the first week, 96 people from 25 countries made trips, including travelers in Brunei, Bahrain and Hungary. That's when it stopped being a learning project. Most of the work since then has gone into the system around the model: - Every place in your day is checked against real venue data: is it still open, and what are its hours that day. - Drive and walking times between stops are computed, not guessed. - Each itinerary runs through 46 quality checks. A day that fails one goes back for repair. Days that zigzag across town dropped from 17% to 10%. - Generation went from 13 minutes to under 2. How it works: answer 8 tap questions, or name where you're going. You get a day-by-day plan with maps, a food guide and a packing list. The first two days are free, and $9.99 unlocks the full trip. I'm here for Hypership all day. Tell me what's missing or broken in your trip. I'll build the best requests live and post each one in this thread as it ships.

MonkeyEatingMango

Tap 8 answers, get a day-by-day trip with real venues

Tap 8 answers, get a day-by-day trip with real venues