Rabbit r1 Review

Initial Impression

I recently got my hands on the Rabbit r1. It’s getting a lot of negative reviews right now for being incomplete, with a very limited set of features. Unfortunately, this is true. However, it has a few good features and a lot of potential for the future.

Build Quality: The Rabbit r1 feels incredibly well-made. It’s solid and durable, which is reassuring. The scroll wheel is a cute touch, but not very practical. I would prefer to swipe the touch screen, which, disappointingly, is disabled.

Services and Subscriptions: The device supports only a few services, none of which I use. It relies on MidJourney for image generation. MidJourney isn’t free, and I’m already paying for too many AI services. This kind of integration with third party services is supposed to be the core offering from this device. Considering there is no developer support, or any way to add your own custom actions, leaves it lacking in reasons to pick it up and use it.

A Few Good Features

One key feature Rabbit showcases is in asking the device questions. I think this is a mistake. An LLM is not a repository of information. The r1 has the ability to query services like Perplexity and Wolfram Alpha, but leaving the interaction open-ended leads to mistakes. Unless the user specifically instructs it to use one of these services, it’s not clear whether a response is coming from those sources or something the LLM made up. Despite its shortcomings, the Rabbit r1 has some promising features:

Voice Recording: The voice recording and transcription feature is excellent. It transcribes long-form dictation accurately and removes most of the filler words and hesitations. While phone apps offer similar functionality, I find the phone distracting due to notifications. The Rabbit r1, as a dedicated button for this task, is a significant advantage.

Potential in Handwriting Conversion: The device has potential in converting handwritten notes, although it currently needs improvement. For instance, it failed to accurately transcribe a page from my work notebook. When I used GPT-4o to convert the same image, the results were flawless. I’m hopeful for future upgrades in this area.

The Biggest Missing Feature

The biggest drawback for me is the lack of developer access. It’s disappointing that they announced this back in January and have had no updates on the subject. I know that they are working on some sort of no-code solution, which might be useful. But developer access would really open up the would give the opportunity for hobbyist devs to really personalize the device into services that they use.

What I’m looking for is something like a web form on my account page. It would have the ability to add an api url, an example of the data to send to that url, and a system prompt instructing the device how to manipulate the sent data. Queries referencing that custom action would use the system prompt from the form to generate an api call and send it out.

This feature would significantly increase the device’s value. Allowing other developers to integrate with the device would offer unlimited growth potential. For devs like me it would be a great hobby device to work with.

Final Thoughts

The Rabbit r1 is not for everyone, at least not yet. In its current state, most people would find it useless. However, I didn’t buy it for its utility. I believe in the potential of LLMs (Large Language Models) as the future of computer interfaces. For me, the $200 investment is more about supporting this concept than the device itself. For the tasks I’m using it for now, I find it worth at least a third of the price. If the company adds the improvements I mentioned, for me, it would become priceless.