Rabbit R1

The Promise

Rabbit burst onto the scene at CES 2024 with founder Jesse Lyu’s presentation of the R1—a bright orange gadget that looked like a chunky Tamagotchi and promised to revolutionize how we interact with apps. The pitch: instead of tapping through dozens of apps, you’d tell the R1 what you wanted, and its “Large Action Model” (LAM) would do the work for you.

The LAM was the key innovation, supposedly. Unlike large language models that just generate text, the Large Action Model would actually take actions—booking flights, ordering food, playing music, anything you could do in an app. The R1 would learn by watching how you used apps and then replicate those actions autonomously.

At $199 with no subscription, the R1 was positioned as the accessible alternative to the Humane AI Pin. The retro design, created by Teenage Engineering, was deliberately playful. The message was clear: AI assistants don’t have to be serious or expensive. They can be fun.

The Rise

The CES launch was a sensation. Pre-orders opened immediately after Lyu’s presentation, and the company claimed to sell 10,000 units in the first 24 hours. By the time shipping began in April 2024, Rabbit had reportedly sold over 100,000 R1s—a remarkable number for a device from an unknown startup.

The hype was fueled by Lyu’s charismatic presentation style and the promise of the LAM. Tech media covered the R1 extensively. YouTubers signed up for review units. The orange device became one of the most anticipated gadgets of 2024.

Rabbit had raised $30 million from investors including Khosla Ventures. The company was valued at over $500 million on paper. For a moment, it looked like the AI hardware category might actually have a winner.

The Fall

Then the R1 shipped, and the story unraveled with remarkable speed.

Android Authority discovered that the R1’s software could be installed as an APK file on any Android phone. The “revolutionary” device was essentially a custom Android launcher talking to cloud services. The “rabbit OS” that Lyu had touted was AOSP (Android Open Source Project) with modifications. Users could sideload the Rabbit app onto their existing phones and get most of the same functionality.

Rabbit pushed back, arguing that the LAM ran in the cloud and couldn’t be replicated with just the app. But this defense highlighted another problem: the LAM wasn’t actually doing what was promised. Reviewers found that most of the R1’s capabilities were just API calls—not an AI “watching” how you use apps and learning to replicate your actions.

The reviews were brutal. The device was slow. The voice recognition was unreliable. The features were limited. The battery died quickly. The screen was hard to see in sunlight. Everything the R1 could do, a smartphone could do better with free apps.

The fundamental promise—an AI that could autonomously control any app—turned out to be science fiction. What Rabbit actually shipped was a clunky voice assistant with fewer capabilities than Siri or Google Assistant, in a separate device you had to carry around.

Warning Signs

• Vaporware demos: The CES presentation showed capabilities that didn’t exist in the shipping product
• “Large Action Model” hype: The LAM terminology was designed to sound impressive but obscured that the technology was just scripts and APIs
• Standalone hardware trap: Requiring users to carry a separate device for functionality their phone already had was an impossible sell
• Too cheap to be revolutionary: At $199 with no subscription, the economics of genuine AI innovation didn’t pencil
• Founder’s background: Lyu’s previous ventures included a failed NFT project—a pattern of chasing hype cycles

Epitaph

🪦 LAM turned out to be LAME