The places are easy. The day is the hard part.
I know what I want to see. I just can't make it into a day.
Alveo turns a scattered pile of wants — a few museums, some restaurants, a market, maybe shopping — into a day that actually holds: paced, ordered around opening hours, and honest with you when you're trying to cram in too much. Type one sentence or hand over your whole list; it meets you at either end.
Alveo — Latin: the riverbed. The channel that gives water its course.
AI tools moved fast on exploration and prototyping. The research, problem framing, information architecture, interaction model, design system, and every final call are mine.
Alveo is a companion for the messiest part of a trip — the middle, where you know roughly what you want but not how it becomes a day. Type "Sofia, four nights, good route" and get a clean, paced plan to use as-is. Or hand over the hotel, the three museums, and the maybe-shopping, and keep pulling threads until every day is dialed. Same system, two depths. Either way, the plan isn't a list of places — it's a course you flow through. Which is the whole reason it's named for a riverbed.
One box. A sentence or a whole pile of wants.
Pace, mornings, food — and what to do when a day runs short.
Paced, ordered around opening hours, honest about too much.
The places are easy. The day is the hard part.
Every trip is two things: the places, and the day that strings them together. The places are everywhere — a friend's list, a saved reel, a thousand dropped pins. The day comes with no instructions. A city hands you ten thousand things to do and not one word on how they become a Tuesday.
So you build the day yourself, and hope. It looks great on the screen — right up until the museum is closed on Mondays, or you've walked eleven kilometers by 2pm, or lunch is an hour across town from everything else. The plan didn't survive contact with the actual city. You don't forget the afternoon it fell apart.
Plenty of tools promise to fix this. They split into two camps — and each one breaks in its own way.
The organizers can't think.
Wanderlog and its kind nail the raw data: real maps, real opening hours, real travel times. Then they hand you a blank grid and make you do all the thinking. Their route optimizer is built to save gas on a road trip — it'll shorten your driving, but it won't put the cathedral first because it closes at four, won't flag that you've stacked eleven hours into one day, won't explain a single choice it made. It's a calculator. Precise, powerful, and completely without judgment.
The generators can't cope with reality.
The new AI planners go the other way: type a sentence, get a full day in seconds. It feels like magic — until you read it. They pack five museums into one afternoon as if teleportation were free, bounce you back and forth across the city, and drop you at a landmark minutes before it closes. One analysis found errors in nine of ten AI itineraries. They're fluent, confident, and quietly wrong — and worse, they can't tell you why they chose anything, or admit when they're guessing.
Credit where it's due — and the seam nobody sits in.
Both camps earned their users. Wanderlog is a genuinely great organizer; the AI planners cracked something hard, turning a plain sentence into a plausible plan. Real credit. But talking to travelers — and reading these tools' own reviews — the same three failures surface again and again, on both sides:
They overpack. Almost nothing protects you from a day that turns into a death march.
They ignore the clock. Opening hours are an afterthought, so the plan marches you to closed doors.
They won't show their work. You get a plan; you never get the reasoning — so you can't trust it, and you can't steer it.
One camp has the data and no judgment. The other has judgment and no grip on the ground. The seam between them — a plan that reasons and holds up — is sitting empty.
What would this feel like if the plan always held — ordered around the clock, honest when you're asking too much of a day, and able to tell you exactly why?
No two people plan alike.
Every plan breaks the same way.
Seventeen people — friends and friends of friends, real trips, not hypotheticals. Small sample, so I read it as signal, not proof. Three things came back clearly.
Everyone juggles. Nobody complains.
Every single person planned across at least two sources — most across three or four. And yet, asked whether planning feels 'scattered and messy,' only 1 of 17 agreed. The friction is real — people have just stopped noticing it. That reframed the problem: the enemy isn't a messy feeling. The cost is hiding somewhere else.
The plan is where it breaks.
15 of 17 have had a trip plan fail in a concrete way — and among those who'd tried an AI planner, the top complaints weren't hallucinations but plans that felt generic or ignored opening hours. This is the real wound — not the planning, the plan. It looks fine on the screen and falls apart in the street.
15 of 17 have had a plan fail in a concrete way.
couldn't fit in everything they wanted
showed up somewhere closed
packed too much and burned out
+ smaller counts: wasted time backtracking (2), a place farther than expected (2), couldn't find a planned spot (2).
There's no such thing as ‘the traveler.’
Asked how they plan, the group split clean across the spectrum — from 'hand me a plan' to 'I over-plan everything.' No single mode wins. One interface can't serve all of them — unless it has a low floor and a high ceiling. That's why Alveo works at two depths: instant for the hand-it-to-me crowd, deep for the day-by-day planners.
“You plan the whole route yourself — and by the end, half of it you don't even want to do anymore.”
Then I asked what would actually help, and one answer led every other: 10 of 17 gave top marks to 'show me why each place and route was picked' — the highest score in the survey. Close behind: warn me when I'm overpacking, order my day around opening hours, one plan I can use now or expand later. People don't just want a plan. They want a plan they can trust and question.
A plan that thinks — and holds up.
The research pointed at one thing: people don't want another list, and they don't want a black box. They want a plan that reasons like a smart friend and survives contact with the real city. Four principles, each answering a failure the travelers named:
Order the day around reality. Opening hours, travel time, and pace are hard constraints, not afterthoughts. The cathedral that closes at four goes in the morning — automatically. No more closed doors.
Know when it's too much. A day has a limit. When you're cramming, Alveo says so — and offers the tighter version that actually breathes. It protects you from the death march.
Always show the why. Every place and every route comes with its reasoning, in plain language. The single most-wanted feature in the survey. You can trust it because you can see it — and question it.
One plan, two depths. Type a sentence and get a clean plan to use now; or open it up and shape every hour. Low floor, high ceiling — because no two travelers plan alike.
Reason, restraint, transparency, range. That's the plan the incumbents left on the table.
Watch it plan four days in Sofia.
One trip, start to finish. A messy sentence goes in; a day that actually holds comes out. Every screen below is real — built and clickable, not mocked. It starts here.
Say it.
No forms. One box, and you fill it however you actually think. A first-timer types ‘Sofia, 4 nights, good route’ and that’s enough. A planner dumps the whole pile — the hotel, two museums, a food market, a couple of great dinners, maybe some shopping. Same box, same system. The floor is a sentence; the ceiling is everything you know. Nobody has to learn a tool to start.
Answer four questions.
Then Alveo asks — but only four, and only the ones that change the plan. Where you’re staying, because every distance is measured from somewhere. When you’re going, because opening hours are a date, not a fact. What you’re into. And one nobody else asks: if a day runs short, what’s your move? You answer a quiet café — and now Alveo knows how to fill a gap without interrupting you later. That’s contingency, collected up front. Four taps, no typing, and a ‘skip to my plan’ door for anyone who’d rather not.
Watch it think.
The plan takes a few seconds, so Alveo spends them proving it’s working — reading your list, checking opening hours, measuring walking times, pacing the days, leaving room to wander. Three conflicts found before you’ve seen a single stop. Most tools hide this behind a spinner and ask you to trust the output. This is the same promise the survey asked for, made before the plan even lands: show me why.
The day.
Here’s the difference. The museum sits at 9:15 because it opens at 9. The Museum of Socialist Art lands at 3:00 because it closes at 5:30. Every stop carries its reasoning in plain language — on the way, 2 min detour, great dinners on your list — and every walk between them is a real measured distance, not a guess. The 15-minute stretch turns clay, on the timeline and on the map, because a long walk is something you should see before you’re standing in it. And at 5:00 the day just stops, and offers the café you said you’d want. Structure, with room left in it.
And when it’s too much, it says so.
You asked for six stops. Alveo gives you four, moves the shopping to Day 3, and tells you exactly why: six in eight hours means rushing all of them. Then it hands the decision back — put it back, or ask why? It’s not overruling you. It’s doing what a good friend who knows the city does: pushing back once, honestly, and then respecting your call.
Steer it.
Nothing here is locked. Change your hotel and every distance in the trip recalculates from the new address. Jump between days. Drag a stop, drop one, add a place, reshuffle the whole thing — the plan bends and re-solves around whatever you touch. Because a plan you can’t argue with isn’t a plan, it’s a suggestion. Low floor. High ceiling. Same screen.
Honest about what's still thin.
Alveo is a concept. The thinking is real, the research is real, and the screens are built and clickable — but a few things here are argued rather than proven. Naming them is the useful part.
The survey disagreed with me — and that was the useful part.
I went in convinced the pain was that planning feels scattered. The data said no: only 1 of 17 agreed. People juggle six tools and have completely normalized it. That forced a better problem: the friction isn't in the planning, it's in the plan — the day that falls apart in the street. This case study is the second version of my argument — the first one was wrong.
The engine is a promise, not a proof.
Everything Alveo claims — ordering around opening hours, honest walking times, knowing when six stops is too many — rests on a reasoning engine I designed the interface for, not the engine itself. The hard part is real: hours data is messy and often wrong, 'how long is enough at a museum' varies by person, and a plan is only as good as the data underneath it. I designed the cockpit. Building the thing it's flying is a different project.
'Too much' is a judgment, and I picked the number.
The death-march detector is my favorite thing here and the least defensible. Six stops in eight hours is too many for whom? I chose the threshold; a 24-year-old and a family of four would draw it differently. The honest version learns your limit instead of assuming it — from what you actually did on Day 1, not from a rule I wrote. Right now, it's an opinion with good manners.
What I'd change.
I'd stop asking and start watching. The interview is four taps, which is short — but the sharpest version of Alveo doesn't ask you how you like to plan. It notices you dropped two stops on Day 1, and quietly builds Day 2 lighter. The contingency question is a step toward that: collect the fallback once, use it forever. I'd push much harder in that direction — fewer questions, more inference — and let the plan learn the traveler instead of interviewing them.
Alveo started as "an app that plans your trip." It became sharper: only ever hand over a day that actually holds. If I got one thing right, it's that.