Park Autopilot ‚AI Real-time "what to do next" for Disney World

Hey HN,

I built a mobile-first web app that solves a problem I had at Disney World: too many choices, not enough time, constant decision fatigue.

*The problem:* You're in Magic Kingdom with your family. There are 40+ attractions, real-time wait times changing constantly, height restrictions to remember, and you have a mental list of must-dos. You waste time debating what to do next, walking past rides with short waits, and standing in long lines at the wrong times.

*The solution:* Park Autopilot gives you real-time "what to do next" recommendations. You set up your party profile once (2 minutes), pick your must-dos, and the app continuously recommends the optimal next attraction based on: - Current wait times (fetched every few minutes via Queue-Times API) - Your party's height restrictions - Proximity to your location - Time-of-day patterns - What you haven't done yet

*Tech highlights:* - React Router v7 (Remix) on Cloudflare Pages - Cloudflare Workers for API + cron jobs - D1 (SQLite) for park metadata and user data - KV for session storage - Recommendation algorithm with weighted scoring (wait time 35%, priority 25%, distance 20%, time-of-day 10%, popularity 10%) - Hysteresis threshold to prevent recommendation flip-flopping - Magic link authentication (passwordless) - PWA with offline support

*Current state:* - Magic Kingdom only (49 attractions with full metadata) - $4.99 one-time Day Pass per party - Soft-launched last week with small beta group - Processing real-time wait data every 5 minutes via scheduled Workers

*Pricing model:* One-time day pass ($4.99) instead of subscription. Felt cleaner than recurring billing for something you use 1-2 times per year.

*What's interesting:* The recommendation algorithm needs to balance multiple competing factors. For example, a must-do attraction with a 60-minute wait might score higher than a nearby ride with a 10-minute wait, but only if the must-do's wait is significantly lower than its average. I used hysteresis (15-point threshold) to prevent the top recommendation from changing on every screen refresh.

Wait time data is surprisingly noisy - theme parks don't publish official APIs, so I'm using Queue-Times which crowdsources from park-goers. I built a cron worker that fetches every 5 minutes and caches in D1 with timestamps.

*Challenges:* - Height data for all attractions (had to manually compile from Disney's official site) - Handling edge cases (park closures, ride downtime, special events) - Mobile-first design with glass morphism that doesn't tank performance - Making recommendations explainable ("Perfect timing! Wait is 40% lower than average")

*What's next:* - Add other Orlando parks (EPCOT, Hollywood Studios, Animal Kingdom, Universal) - Show time optimization (recommend nearby attractions during parades) - Multi-day pass options - Better offline handling

*Try it:* https://parkautopilot.com/?utm_source=hackernews&utm_medium=...

Would love feedback, especially on: - Recommendation algorithm tuning - Pricing model - Technical architecture (is D1 overkill for this?) - Feature priorities

Built this as a bootstrapped side project after a frustrating family trip last year. Happy to answer questions about the tech, business model, or theme park optimization in general.