Rabbit R1 Was a Failure. Now Rabbit Is Betting on AI Agent Software

Jesse Liu is not like that. I think the Rabbit R1 was a failure.

Rabbit R1 struggled at launch

Liu’s gadget was one of the first specialized AI devices created during the virtual-assistant gold rush. You could talk to the Rabbit R1 and ask it to complete tasks, such as booking an Uber or ordering food through DoorDash.

The R1 was announced to considerable fanfare at CES 2024, but it received scathing reviews. Many of its AI agent features did not work reliably, and WIRED gave the device a rating of 3/10.

Since then, AI assistants have improved significantly, and software agents that complete tasks for users have become one of Silicon Valley’s biggest trends. Now Rabbit is trying to make a comeback with software rather than a new piece of hardware.

Rabbit OS3 brings its AI agent to more devices

On Tuesday, Rabbit announced OS3, a standalone “agent operating system” designed to run across multiple screens. OS3 is also available as a software update for the company’s R1 devices, but you do not need an R1 to use it.

Users can access Rabbit OS3 through a desktop browser, Telegram on a phone, or even iMessage. That is a surprising turn for a company that once championed dedicated handheld hardware over smartphone apps, but the market for AI agent software has matured.

Task-performing AI agents are now everywhere in Silicon Valley. Meta’s new Muse agent, for example, uses virtual machines to shop on a user’s behalf. Amazon has blocked Meta’s Muse from crawling its site.

“We went through all this a year and a half ago,” Liu said with a laugh, recalling the controversy surrounding Rabbit’s use of virtual machines to perform app actions for R1 users. So, was Rabbit simply ahead of its time?

Rabbit says the R1 has been financially successful

Liu said the Rabbit R1 has been successful from a financial perspective, particularly because the company does not charge subscriptions for its AI features, unlike many competitors.

According to Liu, Rabbit has delivered more than 100,000 of the orange-red devices, designed by Teenage Engineering, around the world. The company’s return rate is below 5 percent, and its hardware profit margin is approximately 45 to 50 percent.

Rabbit stayed lean after the R1’s poor reviews

“Internally, from an engineering perspective, we didn’t make any wrong bets,” Liu said. “I think the challenge is that compared to other AI studios that have raised billions of dollars, we have very limited resources, both team-wise and financially.”

Rabbit has raised approximately $60 million in total and currently employs 15 people, according to Liu.

The R1’s poor reviews at launch forced Rabbit to make a difficult decision. Rather than rapidly expanding its team or developing new hardware, the company chose to remain lean. However, Rabbit was under financial pressure this time last year. A planned launch in India was delayed by regulations, and a small number of employees went on strike after claiming they had not been paid.

Liu said he respected the employees’ decision and added that Rabbit has secured more funding and is now on stronger footing. “There’s nothing wrong with the health of the company,” he said.

Rabbit has released about 50 R1 updates

Rabbit focused on software updates designed to improve the R1 experience instead of rushing to develop new hardware. Liu said the company has introduced approximately 50 updates over the past year and a half.

He also pointed to Rabbit’s Discord community, which has more than 12,000 users who provide feedback on new R1 features, including greater hardware openness. Tools such as OpenClaw and Hermes AI Agent can be integrated into the R1 instead of using Rabbit’s own model.

Rabbit’s pivot from a dedicated AI gadget to a cross-platform agent operating system suggests that the company may have learned an important lesson from the R1: the future of AI assistants may depend less on specialized hardware and more on how reliably the software completes tasks wherever users already spend their time.