I had a sneak peek at the device, which may end up being Rabbit’s Cyberdeck, the follow-up to the R1, and while it looks seriously cool, the design is not the wildest thing about it.
Instead, it’s what Rabbit founder Jesse Lyu envisages it being used for. The thing is, it’s so forward-thinking that it’s not only hard to describe, but it’s equally difficult for Rabbit to know when it’s the best time to launch the product.
Welcome to the wild world of AI hardware, where it seems the only limits are self-imposed ones.
What is the Cyberdeck?
A design throwback made for the vibe-coding age
“We took inspiration from the classic Sony Vaio P from the early 2000s,” Lyu said during a conversation over video with Android Police.
I can’t show you the renders I saw because they aren’t the final product, and as with everything, that may change between now and eventual release.
At the moment, it’s similar to Sony’s tiny netbook, but with a more futuristic, minimal, squared-off shape.
It has a mechanical keyboard with hot-swappable switches on the bottom half of the clamshell, and a screen on the other half.
It’ll run Linux and is built to interact with AI agents on the move. Each key will be programmable, and only part of the keyboard can be used if required.
There’s no doubt that if it looks like the render, it’ll be as well-received, hardware-wise, as the R1. However, like the R1, knowing what you can do with the hardware is less straightforward, and it’s still Rabbit’s biggest hurdle.
No more coding
All about the prompts
To understand, at least in part, who the Cyberdeck is for, it’s essential to understand Rabbit.
Lyu wants his AI agents to always work and said he felt weird when they weren’t given enough to do when he was sleeping. This drive to ensure high productivity is what inspired the portable nature of the Cyberdeck.
He also told me he doesn’t think any of the Rabbit team still writes code and instead lets the AI take care of it all.
Instead of code, he and the team spend hours writing, in natural language, documents and manuals for the AI agents to follow, so they get the result they want. It’s why there’s a full keyboard on the Cyberdeck.
Understanding these points is key to understanding what the Cyberdeck will be. It’s a keyboard-equipped, Nintendo Switch-sized machine for vibe-coders.
It’s an evolution of the reinvigorated R1’s most successful recent feature, called Creations, where people can create custom tools and apps by using their voice, along with Rabbit’s new DLAM agent, where you control your computer just by talking to it.
It’s not going to replace the smartphone, but through eager use, it could replace just about every app, service, and subscription software package we use.
Through clever magic behind the scenes, the Cyberdeck’s agents will be able to bypass restrictions that normally stop them from communicating with online services freely, allowing them to do more for you. But only if you tell the AI agents what to do.
The impression I got during my conversation with Lyu was that we’re barely scratching the surface of what AI agents will be able to do, and the Cyberdeck will be a portal into a world where you’ll be able to tell it what you want, and it’ll do it. No coding experience required.
Ending subscription fatigue
Enter bespoke, agent-created tools
If that’s what the Cyberdeck (potentially) does, who is it actually for?
Rabbit is not specifically trying to solve one use case, but trying to change the way that users perform tasks, Lyu said, and he used direct experience to illustrate what this means.
Lyu talked about changes he sees at Rabbit and the companies he works with when it comes to AI adoption.
He talked about how many of the R1’s users are young people still in high school, and how other users have ditched software subscription packages after learning they could build something bespoke using the R1 and AI’s capabilities.
AI is all they want to use. They don’t want to go back to Photoshop, they don’t want to go back to any traditional software, and they understand if you have a good platform with agentic coding, they can build what they want by themselves. They don’t necessarily even need a company to do that for them, they can just build their stuff.
Time is not on Rabbit’s side
A fine line between too early and too late
Big-picture thinking like this is exciting, but the trouble is, it’s also abstract, and the concept likely only stops sounding like science fiction to those who are already deeply embedded in the world of AI, agents, and vibe-coding.
To everyone else, it’ll sound like something they’d never use. Unfortunately, educating the general tech fan on how agentic AI’s potential will be a luxury Rabbit doesn’t have.
In the perfect scenario, we would like to make sure everyone understands it before we ship, but that’s always going to be an imaginary scenario because, in reality, the industry is just moving so fast. 10 years ago, it was okay for a company to take a year and a half to build an app. You can make revision after revision. If you spend a year and a half building anything that’s AI, I guarantee you’ll be three generations behind.
Lyu talked about building apps when the Apple App Store first launched, and how you’d need a team of developers and designers to do it.
It took time and effort at the start, and growth was measured in hundreds of apps over a period of months.
The definition of creation has dramatically changed. Now you can achieve this in minutes, by people who have no idea how to program or design, but they can talk and type, and that’s all it takes.
Such a fast pace causes a problem for the Cyberdeck’s release.
Ship too late, and it’ll be behind the curve, potentially overtaken by whatever Open AI and Jony Ive are creating.
Ship early, and like the R1, there’s a real risk of buyers scratching their heads over what the Cyberdeck is actually for.
It was 90-something days between when we started drawing sketches of the R1 on paper to the first unit we shipped. We move extremely fast. Hopefully [the Cyberdeck’s release] is going to be this year. We’re working on it with rabbit speed.
Learning from the R1’s trials and tribulations
Act like a startup
What’s Rabbit’s plan in the face of all this? Lyu was quite clear:
I think a startup should think like a startup. You shouldn’t think you’re Apple, like you magically know everything and can handle everything. Your only advantage is that your idea is ahead of everyone else and you can move fast. You have to get to know your advantages, and think and move like a startup. So that’s exactly what Rabbit is doing.
Rabbit has also been here before with the R1, which had a difficult start in life, but has become something far greater than initially seemed possible after a wealth of attention and updates since release.
Surely, this has also had an effect on the company and the Cyberdeck?
I think we learned a lot from our own playbook. We had extremely dark and difficult times. We’re not in a position of thinking we’re the same as any of the big companies out there, and we understand how to use it to our advantage. The best thing to do is launch aggressively, move fast, and introduce this idea ahead of time. Then talk to our customers and work with them. That’s the best we can do.
The Rabbit R1 is a very different animal today compared to how it was at launch.
Exactly what the Cyberdeck will be at launch remains to be seen, but it seems like it’ll have to arrive sooner rather than later, if Rabbit hopes to avoid a fast-moving industry changing before it has a chance to capitalize on it.
