The park existed.
Its guests didn't.

For almost the entire time I worked on it, I was not allowed to see the park. It was real by then: forty minutes outside Riyadh, six themed lands and the tallest, fastest and longest roller coaster ever built stood against a cliff face, the flagship of Qiddiya, the 360 square kilometre entertainment city at the centre of Saudi Arabia's Vision 2030. But the site was sealed for physical commissioning. The only people through the gates were the ride engineers and safety testers, stress-testing coasters that ran, over and over, with nobody in the seats. No queue had ever formed, no family had ever crossed the park in the noon heat, and no guest had ever asked it a question. I worked on the guest's day from three thousand kilometres away, through plans, models and the pictures you are about to scroll through.

My job was the phone in the guest's hand. Tickets, the live map, the beacons in the ground, ride booking, the AI concierge: the entire day as a guest would carry it, for the first Six Flags ever built outside North America. And over all of it, one constraint that outranked every other: the gates would open on December 31, 2025, ready or not.

A product normally meets its users long before the day that matters, through betas, soft launches and quiet first weeks. This one would meet its first real guest and its finished park on the same morning.

The park had no guests to observe and there was no comparable product in the country to learn from. So I did what I do best when the answers aren't in any document: I went and found ground truth somewhere else. I downloaded every park app on both stores, the Disney and Universal giants down to the smallest regional water park, and lived in them until I knew exactly where each one made a day easier and where it quietly gave up on its guest. Then I got on planes. I walked theme parks across European cities with my phone out and a notebook: timing entry queues, watching where families stopped dead in the middle of a path, waiting to see what people actually did the moment a ride went down. (Not one ride failed on any of my visits, which somehow only added to the pressure.)

The park I was designing for had never held a crowd. But other parks held one every day, and guests are guests everywhere: heat, hunger, tired children, dropped connections and changed plans are universal. The patterns I carried home became the standard every drawing had to be checked against. With that in the notebook, we wrote complete days, hour by hour, for the guests we knew were coming, and every feature had to earn a minute inside one of those days before it earned a place on the plan.

Three days, written down, became the test every roadmap argument had to pass. It is much harder to fight about feature priority when the question is which minute of whose day it serves.

Losing your inputs does not excuse you from deciding. It changes what a decision has to look like. The park's physical facts were solid; its operational ones did not exist. Every behavioural number we ran on, queue times, crowd flow, demand, dwell, was a model with a date attached, so the product was built to survive its own models: everything physical lived as server-side data rather than code. Names, positions, thresholds, hours, whole rides. When a late construction change moved an entrance or a model missed reality, the correction was a data edit that reached every phone in minutes, not an app release that took weeks.

We didn't run the roadmap as a feature list. We ran it as a timeline of one guest's day, from the sofa to the last ride, and shipped in the order the day happens. Scroll through the family's opening day. The drawings follow, exactly as they were sketched: on paper, before the park had ever run a day.

Strip away the desert and the records, and this was a product problem most teams will eventually meet: shipping into a reality you cannot observe yet, on a date you do not control. What survived contact with opening day was not any single feature. It was the way the work was set up.

The guest's written day settled arguments that job titles couldn't, and kept five systems pointed at one experience. Guesses were treated as inventory: labelled, priced, and kept cheap to change in production, so being wrong was a Tuesday, not a crisis. The failure states were designed before the happy paths were finished, because opening day is exactly when rides go down and networks fall over. And the date held, in public, because everything above existed to protect it.