wolfe with an e's blog
The Rabbit R1 chronicles.
Chapter 15: One small step for a baby, one giant leap for AI-kind.
A couple days ago Rabbit R1 founder Jesse Lyu noted on Discord: “We prefer to take baby steps on the complexity of the OS.” In other words: it has to be right before it goes into the wild.
I was really happy to hear this. Far too often companies rush to launch, leaving out hardware features that quickly become obvious as missing and putting software out there that hasn’t gone through a stringent QA process. The results can be disastrous and even company-killing.
The team at Rabbit is moving blindingly fast, but at the same time there’s a serious focus on making the Rabbit R1“just work” at launch. As a result, Rabbit will basically do everything shown in the January keynote. (That’s with the possible exception of travel-related booking, as some regulatory hurdles are cleared.) So what functions will Rabbit R1 have on day one?
Basic features
- Conversation with LLM
- Up-to-date search with Perplexity
- AI vision (camera)
- Bi-directional translation
- Note taking with AI summary
LAM-powered features
- Music
- Generative AI
- Rideshare
- Food
Maybe it seems like you can do all this with your phone. In truth: you pretty well can. What Rabbit can do at launch is not as revolutionary as how Rabbit will do it. No apps are involved: You just hold a button, tell Rabbit to handle these things and it does.
I’m not going to flesh-out these categories since upgraded function seems to be added to the Rabbit OS daily. That’s unlike your phone where the OS and apps are locked until upgraded by downloading updates which can take days just to get approval in app stores. One advantage of the appless Rabbit OS is that the heavy lifting is all done by your private virtual machine in the cloud and updates to your device will be minimal and over-the-air. The improvements will not only be coming fast, they’ll also often be seamless and invisible to the user. So your Rabbit will just keep getting better without you needing to do a thing.
I will mention the LAM features. Some are skeptical if Rabbit’s Large Action Model actually works. While it’s a sparkling-new concept it was conceived 10 years ago and has been in development for four years now. While other AI-based companies were building Large Language Models focused on telling you things, Rabbit was already building LAM that’s focused doing things. By partnering with LLMs to do their thing, Rabbit was already building the next stage of AI evolution: a practical device with the ability to act on your behalf. And LAM is unique enough that Rabbit received a patent for it.
Rabbit will work with agents, essentially automated workflows. But you won’t be required to have an individual agent for every single service you want to connect to. From what I understand, Rabbit will instead work with service types, as listed in the LAM-powered features above. Let’s take the rideshare function. This is a genre that’s likely to work with Uber, Lyft, whoever, since the information needed from these services is the same. Each genre will be able to have multiple agents, so you can connect both Uber and Lyft. And if you don’t use rideshare, you don’t even need to turn this genre on in your Rabbit Hole. Over time, more rideshare services will be added. But more importantly more genres will be added as well.
If we must compare Rabbit to our phone, we have to understand the significance Rabbit brings to the game. It’s not just a new product and new OS, it’s a whole new way of doing things. Rabbit will get new abilities added quickly and you’ll hear a lot less “Oh, my phone can already do that” and a lot more “Oh, my Rabbit can already do that”.
Bespite the baby steps, like with any launch, expect some hiccups. But the philosophy of wanting-to-get-it-right likely means bugs, crashes, things-just-not-working will all be minimal. And those that do happen will be corrected at the speed of Rabbit.
Read Chapter 16: Headed to NYC
