Matthew Berman Demonstrates the Rabbit R1 Pocket Assistant

Rabbit sent me an R1 early and sponsored this video so I can show how I have been using it in my daily life. I also plan separate, unsponsored videos with a full review and a tutorial. My main use so far is as a portable source of answers that I can ask questions wherever I am.

In my experience, the voice interface feels significantly faster than Siri or typing a search into Google. I carry the R1 in my pocket with a mobile SIM for internet access. That is my personal assessment of convenience and speed, not a controlled comparison of the services.

One obvious use for me is in the car, where I cannot conveniently type or use a visual interface. I describe pressing the button and asking questions by voice. I say the search is powered by Perplexity and can return current information, though a connected search does not guarantee an accurate answer.

At a Passover gathering, I asked the device about the holiday and the meaning of the food on the plate. I put it down so everyone at the table could listen. My young son also enjoyed asking repeated questions about Pokémon and their abilities without needing to type a search.

The R1 has two physical controls I use: a gray side button and a scroll wheel. A single press wakes the device; holding the side button makes it listen, and releasing it sends the question. I ask which famous yellow Pokémon looks a little like a cat or rabbit, and it answers Pikachu.

The response feels surprisingly quick to me, particularly given my concern that other AI devices had been slow. I then try a similar question with Siri on my iPhone; Siri says it does not understand. This is one informal comparison, not evidence of general performance across questions.

I ask the R1, “What’s another really famous one?” and it responds with Charizard. For me, the ability to follow the prior question makes the exchange feel conversational. The example illustrates a short follow-up, not a comprehensive test of conversational memory.

I also ask about an upcoming UFC event to demonstrate what I describe as access to current information. The device gives a detailed response, but the video does not independently check the event details; a fluent, web-connected answer should not be assumed correct simply because it sounds current.

My second use case is recording in-person meetings. I ask the R1 to start recording, set it down, and then stop the session using the scroll wheel and side button. I say the recording and an AI-generated summary are sent to Rabbit Hole, the product’s web interface.

In the Rabbit Hole journal, I open a recording from earlier and show its audio player and summary. The new recording also appears there; I explain that a longer recording can produce a summary of the meeting. The demonstration shows the workflow, not an independent assessment of transcription or summary accuracy.

I leave other features for the planned tutorial and review, then recall Rabbit’s earlier launch demonstration and the strong response it received. I say the first 10,000 devices sold out quickly and that another 100,000 were sold over roughly the next week; those figures are my account in the 2024 video.

At the time, I describe the R1 as costing $199 with no subscription fee and say it includes Perplexity and ElevenLabs, with connections such as DoorDash, Midjourney, and Uber. I say devices are beginning to ship and mention a New York pickup event. These are historical launch claims, not a guarantee of current pricing, integrations, or availability.

As someone who likes building products, I appreciate a company trying a different kind of device. I congratulate Rabbit and thank it again for sponsoring the video.