After a series of setbacks, Rabbit is finally showing off the AI technology it promised but failed to deliver with its Rabbit R1.
The company recently posted a video and blog detailing a generalist Android AI agent, showcasing its ability to control apps on a tablet in ways the R1 could not.
The demonstration highlights an agent that handles tasks like finding YouTube videos, gathering ingredients for a recipe, and even adding items to a grocery list in Google Keep.
A Glimpse Into the AI’s Potential
In the demo, engineers use a laptop to type commands that the AI agent then translates into actions on an Android tablet.
The tasks are varied—from searching for a whiskey cocktail recipe to downloading and playing the game 2048.
The AI shows a mixed performance, with some tasks completed smoothly and others stumbling, like when it sends a poem over WhatsApp one message at a time instead of in a single block.
Despite these quirks, the demonstration suggests Rabbit’s AI agent is progressing—though it’s still far from perfect.
One of the engineers even wonders if they should have added line breaks in their prompt to improve performance, hinting that more user input and fine-tuning are needed.
Rabbit’s Ongoing Development
The AI agent showcased in the blog post is a direct evolution of Rabbit’s LAM Playground, a “generalist web agent” launched last year.
The company emphasized that this Android demo is just the core action loop the agent performs, and more cross-platform multi-agent systems are on the way in the coming weeks.
For Rabbit, the unveiling marks another attempt to build momentum after the underwhelming launch of the R1, which failed to live up to founder Jesse Lyu’s ambitious promises.
As Rabbit continues to improve and refine its AI capabilities, it’s clear the company is focused on making up for lost time and delivering a product that works—no matter how long it takes.
Leave a comment