Rabbit R1 review: A $199 AI toy that fails at almost everything
I hate the Rabbit R1. It's yet another sign that standalone AI gadgets, like the Humane AI Pin, are fundamentally useless devices meant to attract hype and VC funding without benefitting users at all. It's like trying to build a skyscraper on quicksand: Today's AI models are great for parlor tricks, but they're ultimately untrustworthy. How do you create a device around that? The Rabbit R1's big selling point has been its "large action model," or LAM, which can supposedly understand what you say and get things done. But really, that's just marketing speak. At the moment, the R1 can barely do anything as an AI assistant. And the few tasks it can actually accomplish, like placing DoorDash orders, are faster and easier to tackle on your phone. You know, the device we already own that can tap into AI features and fast cellular networking. Rabbit R1: design and build I'll admit, the Rabbit R1 looks adorable, but that's mostly down to the design magic of Teenage Engineering, a company that can make a simple tripod look desirable. The R1 is clearly building on the Playdate, another tiny square gadget from Teenage Engineering. Instead of that game handheld’s iconic crank, the R1 has a far less satisfying scroll wheel. Its glossy plastic case also feels a lot cheaper and thicker than the Playdate, almost like what you'd expect from a child's toy. Alongside the dull 2.9-inch screen, there's a unique 8-megapixel "360 eye" camera, which can rotate either towards you or away from you. It's an interesting way to avoid bundling two separate cameras, so I'll give Rabbit credit for that. But the 360 eye isn't meant for taking photos: Instead, it's all about computer vision. You can ask the R1 to describe what's in front of you, from objects to documents and articles, and wait for an AI-generated summary. While this is something that could be useful for people with visual impairments, those users could do the same with ChatGPT, Microsoft's Copilot or built-in tools on their phones (which also have vastly superior cameras). Using the Rabbit R1 is an exercise in futility Beyond its looks, the Rabbit R1 is mostly a failure. Once it’s turned on, you should be able to hit the push to talk button on its side and ask the AI assistant whatever you want: the weather, local traffic or a summary of a recent book. In my testing, though, the R1 would often deliver the weather when I asked for traffic, and sometimes it would hear my request and simply do nothing. The R1 becomes more frustrating the more you use it: Its scroll wheel is the only way to interact with its interface (even though the display is also a touchscreen), and it's simply awkward to use. There's no rhyme or reason for how long you need to scroll to move between menu options. The mere act of selecting things is a pain, since the confirmation button is on the right side of the R1. That button would be far easier to hit somewhere below the scroll wheel — or better yet, just let me use the damn touchscreen! Photo by Devindra Hardawar/Engadget Oddly, the Rabbit's touchscreen does recognize taps whenever you need to enter text like a Wi-Fi network password. But even that process is annoying, since it involves turning the R1 on its side and typing on a laughably tiny keyboard. Honestly, I felt like I was being punked every time I had to use it. (Cue the obligatory, "What is this, a keyboard for ants?") Third-party apps on the Rabbit R1 The more I used the Rabbit R1, the more I felt like it was purposefully designed to drive me insane. It can play music from Spotify (if you have a paid subscription), but what's the point of doing that with its terrible 2-watt speaker? Are you expected to connect Bluetooth headphones? You can ask the R1 to generate art via Midjourney AI (again, with a paid account), but it often failed to show me the pictures that were created. On the rare occasion they did show up, I couldn't actually do anything with the AI pictures from the R1. I'd have to load up Midjourney's Discord server on my phone or computer to share them around. Photo by Devindra Hardawar/Engadget When I asked the R1 to find me an Uber to a local theater, it told me that the Uber service may be slow to load on RabbitOS and isn't available everywhere (uh, thanks?). After 30 seconds of idling, it said the Uber service may be under maintenance, or there may be an issue with my credentials. (I logged out and back into Uber on the "Rabbit Hole" website, which you use to manage the R1, but the error persisted.) “LAM works by operating the Uber web app on the cloud on your behalf,” Rabbit representative Ryan Fenwick told me over e-mail when I asked why I couldn’t get the Uber service to work. “Uber ultimately decides how and whether it serves users, so depending on factors like the location you’re booking from, your ride history, etc., it may vary from time to time. We’re implementing measures that help to improve the success rate and transparency of ride booking through R1, so over time
I hate the Rabbit R1. It's yet another sign that standalone AI gadgets, like the Humane AI Pin, are fundamentally useless devices meant to attract hype and VC funding without benefitting users at all. It's like trying to build a skyscraper on quicksand: Today's AI models are great for parlor tricks, but they're ultimately untrustworthy. How do you create a device around that?
The Rabbit R1's big selling point has been its "large action model," or LAM, which can supposedly understand what you say and get things done. But really, that's just marketing speak. At the moment, the R1 can barely do anything as an AI assistant. And the few tasks it can actually accomplish, like placing DoorDash orders, are faster and easier to tackle on your phone. You know, the device we already own that can tap into AI features and fast cellular networking.
Rabbit R1: design and build
I'll admit, the Rabbit R1 looks adorable, but that's mostly down to the design magic of Teenage Engineering, a company that can make a simple tripod look desirable. The R1 is clearly building on the Playdate, another tiny square gadget from Teenage Engineering. Instead of that game handheld’s iconic crank, the R1 has a far less satisfying scroll wheel. Its glossy plastic case also feels a lot cheaper and thicker than the Playdate, almost like what you'd expect from a child's toy.
Alongside the dull 2.9-inch screen, there's a unique 8-megapixel "360 eye" camera, which can rotate either towards you or away from you. It's an interesting way to avoid bundling two separate cameras, so I'll give Rabbit credit for that. But the 360 eye isn't meant for taking photos: Instead, it's all about computer vision. You can ask the R1 to describe what's in front of you, from objects to documents and articles, and wait for an AI-generated summary. While this is something that could be useful for people with visual impairments, those users could do the same with ChatGPT, Microsoft's Copilot or built-in tools on their phones (which also have vastly superior cameras).
Using the Rabbit R1 is an exercise in futility
Beyond its looks, the Rabbit R1 is mostly a failure. Once it’s turned on, you should be able to hit the push to talk button on its side and ask the AI assistant whatever you want: the weather, local traffic or a summary of a recent book. In my testing, though, the R1 would often deliver the weather when I asked for traffic, and sometimes it would hear my request and simply do nothing.
The R1 becomes more frustrating the more you use it: Its scroll wheel is the only way to interact with its interface (even though the display is also a touchscreen), and it's simply awkward to use. There's no rhyme or reason for how long you need to scroll to move between menu options. The mere act of selecting things is a pain, since the confirmation button is on the right side of the R1. That button would be far easier to hit somewhere below the scroll wheel — or better yet, just let me use the damn touchscreen!
Oddly, the Rabbit's touchscreen does recognize taps whenever you need to enter text like a Wi-Fi network password. But even that process is annoying, since it involves turning the R1 on its side and typing on a laughably tiny keyboard. Honestly, I felt like I was being punked every time I had to use it. (Cue the obligatory, "What is this, a keyboard for ants?")
Third-party apps on the Rabbit R1
The more I used the Rabbit R1, the more I felt like it was purposefully designed to drive me insane. It can play music from Spotify (if you have a paid subscription), but what's the point of doing that with its terrible 2-watt speaker? Are you expected to connect Bluetooth headphones? You can ask the R1 to generate art via Midjourney AI (again, with a paid account), but it often failed to show me the pictures that were created. On the rare occasion they did show up, I couldn't actually do anything with the AI pictures from the R1. I'd have to load up Midjourney's Discord server on my phone or computer to share them around.
When I asked the R1 to find me an Uber to a local theater, it told me that the Uber service may be slow to load on RabbitOS and isn't available everywhere (uh, thanks?). After 30 seconds of idling, it said the Uber service may be under maintenance, or there may be an issue with my credentials. (I logged out and back into Uber on the "Rabbit Hole" website, which you use to manage the R1, but the error persisted.)
“LAM works by operating the Uber web app on the cloud on your behalf,” Rabbit representative Ryan Fenwick told me over e-mail when I asked why I couldn’t get the Uber service to work. “Uber ultimately decides how and whether it serves users, so depending on factors like the location you’re booking from, your ride history, etc., it may vary from time to time. We’re implementing measures that help to improve the success rate and transparency of ride booking through R1, so over time the experience should improve.”
At least the Rabbit R1 was able to get me a sandwich. I asked it to find some lunch nearby and it spent an entire minute communing with Postmates and its AI cloud — the precise amount of time it would take me to complete a GrubHub order on my phone. The R1 eventually returned with three chaotic choices: Subway, a nearby Henri’s Bakery and a restaurant five miles away I’ve never heard of
I opted for Henri’s (they do make killer sandwiches), and the R1 showed me a whopping six menu items. Its tiny screen could only hold a picture of the item, its name and the price — you can't tap into it to get a longer description or customize anything. You can only add items to your cart or remove them. I chose two sandwiches and, to my surprise, the R1 completed the order without ever confirming my payment information or delivery address. It was working entirely off of my DoorDash defaults, and thankfully those were up to date.
As soon as the order was placed, my iPhone started lighting up with all sorts of useful information from DoorDash. I received a confirmation from the restaurant, a detailed look at the bill (the R1 apparently added my default 20% tip) and the name of my delivery driver. It took the R1 several minutes before it confirmed the order, and it only occasionally updated me that it was coming closer.
My sandwiches eventually arrived, but I was more struck by the many ways things could have gone wrong. This isn't 1999; I'm no longer impressed by simply being able to order food online like I did from Kozmo.com (RIP). But even back then, I was able to get a full look at menus and customize things. The fact that I could look over at my phone and see the DoorDash app being far more useful made me instantly lose faith in the R1.
There are other things the R1 can do, like recording and summarizing meetings. But that’s also something several apps can do on my phone and computer. The on-demand translation feature seemed to work fine converting English to Spanish and Japanese, but it’s no better than Google Translate or ChatGPT on my phone.
What’s the point of the Rabbit R1?
All of this leads me to ask: What's the point of the Rabbit R1, really? it certainly can't replace your phone, since it can't make calls or send texts. While you can add a SIM card for always-on connectivity, that just makes it more expensive. It'll still be useless on the go, anyway. Perhaps, you could argue, it's a companion device to help avoid being distracted by your phone. But it's so slow and hard to use that I find my smartphone's notification-filled hellscape far more calming. There's nothing zen at all about having yet another device that you have to buy, charge and carry.
And if you suffer battery life anxiety, you absolutely should stay away from the Rabbit R1. When I first received it, the R1 would burn through its battery while sitting idle, doing absolutely nothing, for eight hours. The first major RabbitOS update helped considerably, but the R1 still can’t last an entire day on a single charge. For a device that has such a tiny screen and offloads its work to the cloud, that’s simply inexcusable.
I suppose you could argue that the $199 Rabbit R1 is a good deal compared to the $699 Humane AI Pin (which also requires a $24 monthly subscription), but that’s like saying rabbit droppings don’t smell bad compared to dog poop. Technically true! But in the end it’s all still shit. The Humane’s projection screen is at least an interesting twist on mobile UI, and its potentially less cumbersome as a wearable. The Rabbit AI assistant, on the other hand, is basically just a chunkier and dumber phone.
Don’t buy the R1. Even if Rabbit somehow manages to deliver on some of the promises of its LAM – like the ability to train the R1 to handle the variety of tasks – I have no faith that it’ll actually work well. My advice extends to every standalone AI gadget: Just stay away. Your phone is enough.This article originally appeared on Engadget at https://www.engadget.com/rabbit-r1-review-a-199-ai-toy-that-fails-at-almost-everything-161043050.html?src=rss
What's Your Reaction?
