Tripeaz

The planner shipped, people started using it, and the obvious next move was to keep adding to it. But instead, I paused to work out what this thing could evolve into. Not a roadmap or the whole product, but where the product fits into a vision.

I started with the thing nobody has fixed. After twenty years of travel products, planning a trip still takes twenty tabs. That isn't a features problem. Every one of those tabs serves a purpose for the user.

The underlying reason is trust. Nothing has earned the right to decide for you. So you open more tabs with more products. Every travel product in the category answers that by offering more. More options, more filters, more generated itineraries. More is what created the problem.

Earning that right is the whole product problem, and you can't do it in one screen. It takes time to learn someone's preferences when they travel. Everything below is the thinking that came out of stepping back, and it's what drives what gets built next.

The process of taking a trip breaks down into four jobs, and they require different things from the design. Dreaming tickles curiosity and catches sparks. Planning narrows toward a decision. Being there should stay quiet until the moment it's needed. Coming home closes the loop and starts the next one.

Each phase produces a different kind of evidence about what a person likes. Dreaming produces implicit signal, the stuff people don't know they're telling you. Planning produces explicit signal, the choices and the rejections. The trip itself produces real-world signal. Coming home produces reflection.

Point all four at one model of a person's taste and the phases stop being a single funnel. They become a loop that's personalized to the individual.

Tripeaz started at Plan. It's the phase I understood best, but Dream felt like the next feature to build. But Dream is a different design problem, not an extension of the planner.

Nobody starts by planning the logistics. They start with a clip someone sent them, or a place that keeps coming up. They're not trying to go anywhere yet. Thoughts are just percolating.

Dream's job is to observe taste and feed the model. The user should feel like they're simply exploring the world. The longer someone spends dreaming, the more of the itinerary can be proposed instead of asked for.

Filling up an empty itinerary manually is tedious, and most people won't do it. They open the planner, look at the blank canvas, and leave. Dream's goal is to solve the blank canvas problem.

The Dream prototype is the first attempt at that problem. It runs and lives on an unmerged branch, so nobody using Tripeaz today has seen it.

There's no search box. Every destination is a star, and you move through a map of them with previews playing under the cursor as you go. Catch the ones you like. When you're ready, what you caught becomes a trip.

The current iteration of Dream runs on a multi-stage pipeline. When it collects enough signals about your next trip, it writes themed searches for districts in the city, pulls licensed footage against them, checks every clip with vision before accepting it, then plans the stops and renders a vertical montage.

A button at the end of that film turns what you watched into an itinerary. That button is the seam between the first two phases.

The planner is working. Live and Remember exist only in the diagram. Powering the full Dream pipeline with AI and releasing it broadly will cost real money and will likely require funding. The most practical thing in the near term is tackling growth, to show engagement and retention.

Sample itineraries are the interim answer to the blank canvas. Start from something instead of nothing, then edit it into your own trip. That's what the growth work is actually for, and the goal is to turn that pipeline into something anyone can point at their own destination.

Dream is where it's going. What you looked at, lingered on, and skipped is already a description of the trip you want. Get that right and the canvas is never blank, because by the time you reach the planner you've already told the product plenty.