The Rabbit R1

When I ordered the Rabbit R1, the ‘AI Pocket Companion’, back in January 2024, I fully expected to be writing a post later in the year declaring it the worst bit of tech I’ve ever bought. The price point, unlike the Humane Pin, leant itself to this kind of frivolous purchase as well, so I took a punt. Come June 2024, when it finally arrived, I was not disappointed. This device is, indeed, the worst bit of tech I’ve bought this year, perhaps ever. The story, however, does not finish there … 

Hardware

First things first, let me just speak a little about the hardware itself. The Teenage Engineering designed little device is a really nice product. It’s very light, thin, and small, with a decent screen (for what it is). It includes just one physical button on the side, held down to initiate the Rabbit AI functions, along with a scroll wheel built into the side of the device and a tiny little camera with a motor that lets it actually spin around to focus either towards the user, or away from you. It’s such a fun design, with an eye catching bright orange lick of paint. When I received the shipping notification that the R1 was on its way I cleared a space on my shelf between the Analogue Pocket and my Anbernic, but so far I’ve got to admit that the design is so enjoyable it’s remained sitting on my desk since arriving.

Despite the battery being pretty poor, I really can’t fault the hardware. I suppose if I had to be hyper critical I’d say that it does look like it would scratch very easily in a pocket. Seeing as it’s marketed specifically as a pocket companion, that’s not ideal. It does come with a nice transparent plastic case that hinges open to act as a little stand, however, so it does have some provided protection that doesn’t add a lot of bulk. 

Software

When you move away from just looking at the device, and start trying to use the thing, the experience takes a real nose dive sadly. I’ve found that I can get a fairly decent use out of AI for a few work tasks recently, and I’ve experimented in using Perplexity AI and some other tools, but for me AI is still firmly in the novelty category for me. It has not only failed to set my world on fire, it’s barely given me a sweaty crack. It just feels like far too much effort to open Copilot and then trying to think of the very specific wording I need to use to invoke the response I want. The Rabbit R1 takes this frustration and sub par response behaviour to the Nth degree. The AI features you can initiate from the R1 are, even for AI services, pretty limited. They published a Prompts Guide to help with this. It has integrations with Perplexity and Wolfram Alpha for answers to standard questions, which works well. If I want to ask it what the weather is like, however, I have to specify that I want it on Celsius every time, which highlights just how janky all of this is. You can also get it to record a meeting, or some dictation which it will save and then let you reference or ask for it to recall some information from it at a later date. All very basic stuff here really, and as is often the case with AI systems, at least for me, I rarely find this useful. 

There are also some features that take advantage of the little camera, such as allowing you to take photos of things to identify and ask about. This is very deep into novelty territory at this point. The camera quality is also so bad it can struggle to identify things sometimes. 

The point that things get really bad for the R1 is when you start exploring the Rabbit Hole feature. When you ask the device anything, take photos, record anything, or interact in any way, the responses are all stored in the journal on a web portal at hole.rabbit.tech. Cool name aside, this thing is shocking. Firstly, I don’t trust Rabbit enough to have it holding onto recordings of my meetings, and especially any photos. You can delete entries from this portal, but I still would assume they’re holding onto this data and training their models on it. Things get even worse, however, when it comes to connecting the device to the few third party services that you can link up to the Rabbit system. None of the connections use the standard oAuth or API integrations with third party services. They offer an Apple Music connection, for example, so you can play songs from the R1. To connect to this, however, the portal opens a connection to a Virtual Machine that has the web version of Apple Music open, expecting you to enter your full Apple ID and password into it. Yes, really … 

There is no chance in hell I’m going to be putting my full Apple ID credentials into this dodgy system. So no music for me. I don’t even want to put my Discord credentials into this for Midjourney integration. It’s just a shockingly bad implementation that I’m actually shocked didn’t kill this thing instantly. 

As soon as I saw this I joined the Rabbit Discord to see if this was some kind of temporary system and to ask why on earth people were using this in its current state and, as expected, it was full of sycophants and apologists telling me that I was wrong for finding this extremely dodgy. Simon, the Community Manager and Jesse, the CEO, even joined in the conversation and tried to gaslight me into believing that oAuth is an outdated system and dodgy VNC / VM connections was some kind of look into the future. When Simon asked me why I think oAuth is more secure I simply stated that I have never seen another company manage integrations in this manner so you tell me why Rabbit are right and, apparently, the only ones doing it the right way across the entire industry. There was, unsurprisingly, no further replies. This, to me, is completely unacceptable and the R1 deserves to disappear into obscurity for this decision alone. 

And yet … 

All that being said, the R1 is still on my desk and not relegated to the shelf of death so why is that? I’ve stated in the first paragraph that the R1 is the worst tech product I’ve bought, and I 100% stick with that, however, when I adjust my perspective and consider it as a toy, that changes things somewhat. If I take any pretext of this being the future, or a device that is going to add untold levels of productivity to my life, and look at it is a (relatively) cheap and fun toy I start to see a little appeal to it.

Take the camera, for example. Taking photos of things to ask the R1 what it is is slow, and inconvenient. Something else it does, however, is let you take a photo of something and it will quickly generate an AI art interpretation of it. Yes, AI art is bad, yes it won’t help with your productivity, but it is also fun. I’ve found myself shooting photos of all sorts of things just to see what it would create. The example below is a photo of my Optimus Prime LEGO, as an example.

They recently added an integration with Suno, an AI music generator, which I can thankfully log into with a burner account into their crappy VNC. I can now ask the Rabbit to make me songs based on my voice inputs. Again, pointless as a productivity tool, but really fun and my daughter loves it. Even my non-techy wife enjoyed that one. 

If this was a £600 product with a monthly subscription like the Humane AI pin I would, of course, look at this device with a completely different lens. At £200, it’s not exactly cheap, but it’s just on the threshold for me where it being a desk toy, comprising some very cool hardware design and a few fun features, I’m certainly finding the joy in it. I fully expect the servers will not be running far into 2025, if it even makes it that far, but for now I’ve got exactly what I was expecting when I purchased it in January. To Rabbit’s credit, they have released an OTA update every week since release to add new features, so it’s likely improved a fair bit since the initial reviews.  It is both the worst tech product I’ve bought this year, and the best toy to embody 2024 tech trends I could imagine.