The Hype Machine Is Real
When the Rabbit R1 launched in 2024, it was a total flop. The device promised to be an AI assistant that could control your apps, but the software was buggy, the battery died in hours, and reviewers called it a “glorified Tamagotchi.” I was one of those critics. So when Rabbit announced the R2 in early 2026, I was skeptical. But the specs looked promising—better hardware, a new operating system called Rabbit OS, and a partnership with OpenAI for GPT-5 integration. I bought one on launch day. Here’s what happened.
First Impressions: It’s Actually Cute
The Rabbit R2 is about the size of a deck of cards. It has a small touchscreen, a camera, and a scroll wheel. The design is minimalist—black and white with a single button. It feels premium, unlike the plasticky R1. Setup took about 10 minutes. You link it to your phone, log into your accounts (Uber, Spotify, Gmail, etc.), and the device learns your habits. It sounds creepy, but the privacy policy is surprisingly transparent—your data is processed on-device, not in the cloud.
What It Can Do (When It Works)
The R2’s main feature is the “Action Agent.” You can speak natural commands like “Order my usual coffee from Starbucks” or “Find the best route to avoid traffic.” The device then uses its AI to interact with your apps on your behalf. I tested this extensively.
Ordering coffee worked flawlessly. I said, “Get me a medium black coffee from the Starbucks on Main Street.” The R2 opened the Starbucks app, placed the order, and even sent me a notification when it was ready. The whole process took about 30 seconds. I was genuinely impressed.
Uber rides were hit or miss. The R2 could call a car to my current location, but when I tried to schedule a ride for the next day, it got confused and ordered an immediate car instead. I had to cancel and get charged a fee.