With high hopes of finding some hidden gems in our home cities and $100 (£77) each burning a hole in our pockets, we—Natasha Bernal in London and Amanda Hoover in New York—asked AI to plan out the perfect day.
We decided to use Littlefoot, an AI-powered local discovery chatbot that can generate experiences in 161 cities around the world. It was created by Bigfoot, a startup founded by former Airbnb executives Alex Ward, James Robinson, and Shane Lykins that purports to enmesh the minds of GPT-4o and Claude 3 Haiku in addition to 50 information sources such as EventBrite and Google Places.
We told Littlefoot our respective starting points, dates, and times, and introduced some caveats: Amanda asked that her New York tour be dog-friendly; Natasha was obsessed with avoiding London’s crowded tourist hotspots.
The results were, frankly, rather mad. Right now, Littlefoot has no concept of time or space or what a human being might find interesting. Its recommendations vary wildly from the incredibly niche (climbing up a hill in South East London) to the wildly vague (going to the London Zoo, no further instruction provided). The same attractions—such as the London Eye, the Namco Funscape arcade in Romford, a cycling studio in Brooklyn—kept coming up in recommendations, to the point that we suspected it might be paid-for advertising. (Bigfoot has confirmed that is not the case and that it has no plans to offer sponsored picks.)
It recommended back-to-back gym sessions in London, a concert and helicopter tour in New York that were out of our budget, restaurants for lunch that didn’t open until dinner time, and itineraries that would have sent us criss-crossing around our respective cities. In London, Bigfoot’s map function showed two out of the four suggested destinations in completely wrong locations, an issue that the company says it is working on.
In a response to Wired, Bigfoot clarified the incorrect locations were due to errors in the major location APIs that Littlefoot uses, and that the company has been updating the product for a more accurate experience.
“While we expect to face typical challenges associated with an early company, we are confident in our ability to meet them as we acquire more resources and continue to refine our approach based on user feedback,” says Bigfoot CEO Alex Ward. “We’re a preseed startup of six, and itineraries aren’t meant to be perfect just yet. But we are working to do everything we can to get there in the not-too-distant future.”
Bigfoot says its features—which are currently very contingent on the location you provide and how you phrase what you’re looking for—have been tested by 70 to 80 alpha users this year, and the company is refining the platform based on feedback.
A Day Around London’s Queen Elizabeth Olympic Park
I picked a day centered around the 560-acre sporting village, which features pedal boats, a track cycling arena, and tennis courts. I had never been before and assumed it would be great fun. It wasn’t.
My day started at 10 am at WIRED’s office in Central London. The first stop was in East London, to eat at a place called Pizza Union, which didn’t open until 11 and which a TikTok video claimed had slices priced £6. (It was wrong — it only sold whole pies.) Armed with Google and a comrade, fellow Londoner and WIRED staffer Sophie Johal, I marched to the underground for a 3-mile trip to Aldgate East, a place I can confidently say no one goes to voluntarily.
We stalled until the pizza place opened. I ordered a margarita pizza pie and a bottle of water, and we sat on a wooden bench on the sidewalk, complete with a view of a strip club called White’s. I choked a bit on an inordinately crispy piece of basil.
Then we walked just over a mile to the next stop, the historic Borough food market, near London Bridge. On the way, in the 17th-century church of Saint Magnus the Martyr, we found a 2,000-year-old crusty-looking wooden log randomly tucked in under the entrance way. Were it not for this trip, we might have lived our entire lives never having seen it.
Across London Bridge, we arrived at the market, where we had to navigate around bag-wearing people and other obstacles to find Horn OK Please, a random Indian street-food stall that happened to be one of the only vegetarian eateries in the entire place. We stood next to a trash bin at around midday and ate dosa chaat (a wrap filled with masala potatoes, onion, coriander, and chutneys). I was sweating profusely, felt like I had eaten a brick, and began to worry about whether I would make it to the Olympic Village at all.
We jumped back on the tube and headed in a circle to Trafalgar Square—2 miles away and five minutes from WIRED’s office. We stood in a crowded queue for 10 minutes outside the National Gallery, and after seeing some familiar public art from the likes of Degas, Murillo, Raphael, and Gainsborough, rewarded ourselves with treats from the gift shop (some badges featuring Van Gogh and Manet artwork and the museum logo, and a chocolate coin—the only things we could afford on our budget).
Back on the tube, I took the Elizabeth Line—opened in 2022—for my first time ever, to get to Stratford and the Olympic Park. Severely tired and dehydrated, I led us beyond mega-supermall Westfield, which overlooked a gigantic car park, to a not-very-scenic path near blocks of flats. By that point, we had figured out that the pedal boats weren’t working, that you had to take a training course for the cycling track, and that the tennis courts were unaffordable on our meager remaining budget.
There was no sign of the hideous steel Orbit Tower, or anything else noteworthy for that matter—we just saw blocks of midrise flats and a rather large park. Invoking the spirit of the Olympics and taking advantage of the fact that we were practically alone, we attempted to complete an obstacle course in a nearby fitness playground. (We got stuck on the equipment a few times.)
And that’s how we ended up lost near the Wetlands Walk close to the River Lea, cycling around on the only available rental e-bike I could find. Unfortunately it could only travel at breakneck speed, so I spent about 20 minutes screaming in genuine fear as I hurtled back and forth through the same bit of stinky swampland, with Sophie perched precariously on the fender like a human pannier.
I gave up attempting to find anywhere after it started to pelt down rain, and realized I had paid £6 for the e-bike, for the pleasure of going absolutely nowhere. We never even found the bike track, and were too exhausted to go to our final stop, six miles away at Bradley’s Spanish Bar. Instead I had an ice cream (two scoops: one straciatella, the other chocolate and hazelnut) at a place near the station. I had a cuppa, and at £64 spent, called it a day.
In the darkest moments, I would have described the itinerary as a teeny bit evil and a lot pointless. It would take two minutes with an internet connection to come up with a better plan, but it would not be as chaotic and random—and ultimately unforgettable.
As soon as I got home, I asked ChatGPT to come up with a rival day: It suggested a cultural walk along the South Bank, winding around Big Ben and the London Eye, with stops to enjoy some art and culture along the way. It made sense, but a small part of me—the same one that screeched in delight at seeing a giant slide that we could go on—wonders whether there was some magic in Littlefoot’s mundane (and inane) picks.
A Day Around Washington Square Park and Central Park in New York
I wanted to see whether Littlefoot could do something a little challenging: Plan a day around Manhattan that I could bring my dog, a 22-pound poodle mix named Charlie, along on. Turns out, it couldn’t.
The app gave me recommendations for a park in New Jersey that wasn’t accessible without a car and two restaurants with the word “dog” in the name—although I’m not sure dogs would be allowed inside—and repeatedly tried to persuade me to take a spin class in Brooklyn, I’m assuming because the word “friendly” was in the description. It also highlighted an expensive helicopter tour of Manhattan when I asked for outdoor activities and, at one point, a Noah Kahan concert (sold out, with only pricey resale tickets available).
Sifting through the options took some work and reprompting. Some of the restaurants recommended were closed at the times Littlefoot suggested I go, and they would send me between the city’s boroughs, or even further out of my way—even when I asked for a day in Manhattan that was completely accessible by subway.
I combined and pruned two itineraries to save myself some agony of hours on the subway. I left Harlem at around 8 am and headed to my first stop: the Obelisk in Central Park, a 3,500-year-old monument brought over from Egypt. Charlie was particularly happy with this plan, as we made it to the park during off-leash hours, and he was free to roam the greenery and trails with other dogs.
We walked a mile across Manhattan’s Upper East Side to get to Claire’s Kitchen Cafe, a tiny Mediterranean-inspired spot with outdoor seating that Littlefoot had recommended. It definitely wasn’t a tourist trap, but rather a neighborhood spot. I had a cream cheese and smoked salmon breakfast sandwich and a latte, both of which were good, but not necessarily something I would travel to a new neighborhood to eat.
Charlie and I took the subway downtown and got off near Washington Square Park, our next suggested stop. The park was quieter than normal on a weekday morning, although I saw some probable TikTokkers running up to people on benches to interview them. We walked past the iconic archway and fountain, and stopped at the dog park, where I threw a tennis ball for Charlie.
We got back on the subway and rode to the AI-recommended 9/11 Memorial & Museum—a somber stop mixed in with all the recommendations for parks and pubs. With my dog in tow, I couldn’t go into the museum, but we did walk around the memorial, and Charlie made a few friends with tourists there.
Our last stop was at the Dead Rabbit for lunch. It was a quirky Irish pub with upscale bar food and drinks, and a bunch of dollar bills taped to the shelves holding liquor. It’s been deemed the best bar in the world by some, and luckily, it was dog-friendly, with one little table out front that was empty when I arrived. I had a burger and a drink, which ran me about $50 with tip. (Charlie slept at my feet during lunch, undisturbed by nearby construction sounds, but completely knocked out.)
About five hours later, I was exhausted and had spent $90 between the subway and food. We had traveled from Harlem, toward the top of Manhattan, to the Financial District at the bottom, taking five trains and walking 5 miles to hit Littlefoot’s recommended spots. I had to rely on my own knowledge of the city to create the most efficient route between the various recommendations provided by Littlefoot.
Afterward, I ran a similar prompt through Perplexity’s chatbot, and found an itinerary with more details about prices and how to get around. This wasn’t perfect, either. It recommended I take Charlie to Manhattan’s iconic High Line, which doesn’t allow dogs, and gave vague advice about buying a bagel from a street vendor “nearby” Central Park. But it did give recommendations in several known parts of the city that would keep me under budget.
Littlefoot got me to try some new restaurants and explore some new corners of famous parks—plus, it got me out of the house and away from my routine. But had I followed its recommendations exactly, I may have ended up going to Brooklyn or Queens to find restaurants with their doors closed, or gone way over budget on some experiences. AI might have some new recommendations, but until the technology improves, I’d rather just ask a fellow New Yorker where to go.
Updated: 7/22/2024, 11:30 am EDT: Following publication, Bigfoot clarified which specific LLMs that Littlefoot uses to generate experiences.
Correction: 7/22/2024, 4:30 pm EDT: Wired also removed reference to Tripadvisor, which is not one of Littlefoot's information sources.
Correction: 7/23/2024, 10:30 am EDT: Wired corrected that Pizza Union sells pizza by the slice.
