Rabbit Returns: Now Featuring an AI Agent App

Jesse Lyu believes the Rabbit R1 was not a failure.

Lyu’s device was one of the pioneers in the surge of dedicated AI hardware designed to function as a virtual assistant. You could talk to the R1 and request it to perform tasks—like booking an Uber or ordering food through DoorDash. After a highly publicized debut at CES 2024, it faced harsh criticism. The agent functionalities on the R1 simply didn’t perform adequately. WIRED rated it a 3/10.

Since those initial days, AI assistants have seen significant advancements, and the technology is now trending. Consequently, Rabbit is re-entering the spotlight.

On Tuesday, Rabbit unveiled OS3, an independent “agentic operating system” designed to function across various displays. It also serves as a software upgrade for the R1 device, though an R1 is not a prerequisite for using it. Users can access OS3 via a desktop browser, Telegram on mobile, or even iMessage.

This marks a remarkable shift for a company that previously prioritized handheld devices over smartphone applications, yet the market for this type of software has matured.

In Silicon Valley, the buzz is all about agents that complete tasks on your behalf. Meta’s new Muse agent was recently prohibited from using a virtual machine to shop on Amazon for users. Lyu chuckles as he mentions this, saying, “We encountered all of this a year and a half ago,” referencing the R1’s controversial use of a virtual machine for executing app actions on behalf of users. Was Rabbit simply too ahead of its time?

Lyu asserts that from a financial standpoint, the Rabbit R1 was remarkably successful, especially since the company does not charge a subscription for its AI features, unlike numerous competitors. He claims they sold over 100,000 of the stylish orange-red gadgets designed by Teenage Engineering, with a return rate of less than 5 percent, achieving an impressive margin of 45 to 50 percent on the hardware.

Jesse Lyu, CEO and founder of Rabbit.

Courtesy of Rabbit

“From an engineering standpoint, we didn’t make the wrong choice,” Lyu states. “The real difficulty lies in being extremely resource-constrained, both in terms of team size and finances, compared to other AI companies that have raised billions.” He mentions Rabbit has secured around $60 million in total funding and has a team of 15 employees.

However, following disappointing reviews after the R1’s initial launch, the company faced tough decisions. Rabbit opted not to grow its team and maintained a lean structure. Last year at this time was particularly challenging. A planned expansion into India hit regulatory roadblocks, placing significant financial pressure on the company, leading to some employees staging a strike over unpaid wages. Lyu expressed respect for their stance, noting the company has since obtained further funding and is on solid ground. “There are no health concerns regarding the company,” he asserts.

Lyu explains that the team opted against rushing into the development of new hardware and instead concentrated on enhancing software updates for an improved R1 experience. To date, he reports that Rabbit has issued around 50 updates in the last 18 months. He highlights a dedicated Discord community of over 12,000 users offering feedback on potential new features for the R1, such as opening up the hardware. For instance, users can incorporate OpenClaw or Hermes AI agents into the R1 rather than relying solely on Rabbit’s proprietary models.