What we’re building, and why.
An AI trip planner that argues with you when it should.
Fortrip was built for the kind of traveler who already does their homework. Reddit threads, three guidebooks, a friend’s blog from 2019, a half-typed itinerary in Notes, two browser windows on TripAdvisor that no one has the will to close. By the time most of our users open a chat, they don’t need to be told that Lisbon has good food or that Tokyo is busy in April. They need someone to read what they’ve drafted and tell them the parts that won’t survive contact with the city — the 9pm dinner reservation in a town where most kitchens close at 8, the train transfer that looks twenty minutes on a map and turns out to require a sprint between platforms, the optimistic Day 3 with five neighborhoods stacked on top of a flight, the hotel that’s well-reviewed but forty minutes from the part of town the rest of the trip is built around. They’ve already done the inspiration work. What they need is someone to argue with them about the execution.
That’s one kind of user. The product also serves travelers who haven’t done the homework yet — the ones starting from a blank page who need help shaping the brief itself, and the ones planning multi-city trips where the constraints multiply faster than the days fit. The shape of the help shifts. The principle stays: better questions, sharper trade-offs, a trip that holds up.
The problem with most trip planning is not the ideas. People mostly know what kind of trip they want. The problem is the details, and the details are where AI tools have historically been useless: they generate the shapeof an itinerary — three days, neighborhoods bullet-listed, a museum per morning — but they don’t notice when the shape collides with reality. They don’t push back on a route. They don’t say “this is the wrong order, you’ll lose half a day,” and they don’t catch the kind of mistakes that only show up after you’ve booked the train. They certainly don’t tell you that the festival you planned around was already announced for the week after you leave, or that the boutique hotel everyone recommends is on the wrong side of a river you’ll have to cross five times. Fortrip exists because we think a good travel tool should do all of that, and the bar should be higher than what a chatbot trained on travel blogs and review aggregators can offer.
Most trip planning fails in the details, not in the ideas. Good tools work on the details.
How Fortrip started
Fortrip wasn’t started by engineers looking for a vertical to apply AI to. It was started by someone who used to design trips for a living — Boyuan Dong, who spent years building themed itineraries by hand and guiding travelers through them. He’d seen the same handful of failure modes again and again across travelers from very different backgrounds: the optimistic Day 3, the badly-sequenced cities, the hotel that’s 40 minutes from the part of town that actually mattered, the route that looked clever on a map and lost half a day in practice. What nobody had built, as far as he could tell, was a planner that caught those mistakes beforethe trip happened — one that could behave like an experienced guide reading over your shoulder. Fortrip is that planner, built by a team that came at the problem from the other side — product engineering, applied AI, and the systems work underneath — and converged on the same conclusion from a very different starting point.
The product
What sits behind the chat today is three tools, all live. The Planner takes a sparse brief — eight days, food-led, mid-budget, late May, no big museums — and turns it into a day-by-day itinerary you’d actually follow, with reasons attached to every choice. The Validator takes an itinerary you’ve already drafted (yours, a friend’s, one a different AI gave you) and reviews it across feasibility, pacing, sequencing, and personal fit, then tells you what to change. The Transport Optimizer compares flights, trains, buses, and ferries between every stop and points to the option that fits thistrip — fastest is not always best, and cheapest rarely is. The three sit inside the same conversation, so the trip you’re shaping never has to leave one place. The reasoning layer underneath runs on proprietary models we’ve built for travel specifically, not a wrapper on a general-purpose chat with a travel skin.
The team
The team is four people. Boyuan Dong (CEO) is the operator’s voice in the room — he comes from years of designing trips by hand for other travelers. Derrick Jiang (COO) brings the product-and-systems thinking that figures out where something breaks and designs a fix that holds. Shuhao Dong (CTO) owns the model architecture and the reasoning layer behind every recommendation. Bruce Lan handles full-stack engineering and the thousand small details between a model’s answer and a user’s actual trip. The longer bios live on the team page.
How recommendations work
Every recommendation in Fortrip comes with reasoning. We sort hotels for the trip, not for the commission. When we suggest a neighborhood, we say why it fits the way you described the trip, and what trade-off it carries. We think personalization and operational reality are one problem, not two: a trip that fits your taste but breaks on the train timetable isn’t personalized, it’s wishful, and we’d rather sweat both layers than only the one that’s easier to demo. When a booking is better-routed through a specific partner, we say so in plain language rather than ranking it up quietly. We think the planner that earns the user’s trust is the planner that gets booked through — that’s the whole business, and we’d rather get there honestly than dress up a worse recommendation with better marketing. Scarcity and price-trend warnings exist inside our reasoning (“worth locking in this week — this property has been moving fast since the May festival was announced”) but never as red-banner retail urgency, never as countdown timers, never as fake social proof. The voice of a concierge, not the patter of a discount site. We think travelers can tell the difference, and we think the difference is what makes the planner worth using twice.
Today, and what comes next
Right now Fortrip’s strength is the moment between a vague idea and a booked trip — the planning, the validation, the routing. The three tools above are live and improving on a near-weekly cadence as the models behind them get better at travel reasoning specifically rather than chat in general. Over the next year we’re extending that line in both directions: deeper into inspiration on the early end, helping people figure out whether a destination fits them before they invest hours in planning it; and through booking and on-trip support on the other end, so the same conversation that built the itinerary can adjust it when the flight gets delayed, the museum is closed for installation, or the weather turns. The product gets stronger as the trip gets longer.
What this should add up to, if we get it right, is travel with a little more confidence and a little less of the quiet doubt that says you’ve missed something — and more time spent on the parts of planning that are actually fun: the wandering through Wikipedia, the half-formed dinner plan, the friend’s blog from 2019. Curiosity is the reason any of this exists. Fortrip is built to keep it intact.