Jesse Lyu would not assume the Rabbit R1 was a flop.
Lyu’s gadget was among the many first within the gold rush to create devoted AI {hardware} that acted as a digital assistant. You could possibly communicate into the R1 and ask it to finish duties for you—e book an Uber; order meals on DoorDash. After a buzzy launch at CES 2024, it earned scathing evaluations. All of the agentic stuff on the R1 simply did not work that nicely. WIRED gave it a 3/10.
Since these early days, the capabilities of AI assistants have vastly improved, and now the tech is all the fad. So Rabbit is banging the drum once more.
On Tuesday, Rabbit introduced OS3, a standalone “agentic working system” designed to run throughout a number of screens. It additionally arrives as a software program replace for the corporate’s R1 gadget, although you do not want an R1 to make use of it. As a substitute, you’ll be able to entry OS3 by means of a desktop browser, Telegram in your cellphone, and even iMessage.
It’s a placing pivot for an organization that when championed the hand-held {hardware} over smartphone apps, however the setting for such a software program has ripened.
Brokers that deal with duties for you might be all anybody in Silicon Valley talks about. Meta’s new Muse agent was simply banned from crawling Amazon to buy on a consumer’s behalf utilizing a digital machine. Lyu smirks as he brings it up, saying, “We have been by means of all of this a yr and a half in the past,” calling again to the R1’s controversial use of a digital machine to execute app actions in your behalf. So was Rabbit simply too forward of the curve?
Lyu says from a financial perspective, the Rabbit R1 was tremendously profitable, particularly contemplating the corporate would not provide a subscription for its AI options, in contrast to many opponents. He claims the corporate delivered greater than 100,000 of the orange-red, Teenage Engineering-designed devices worldwide, with return charges of lower than 5 %, and earned roughly a forty five to 50 % margin on the {hardware}.
“Internally, from an engineering perspective—we did not make the unsuitable wager,” Lyu says. “I believe the problem is actually being extraordinarily restricted in sources, team-wise and money-wise, in comparison with different AI studios that raised billions of {dollars}.” He says Rabbit has raised round $60 million in complete and at present employs 15 individuals.
However after dismal evaluations upon the R1’s launch, the corporate needed to make robust selections. Rabbit determined in opposition to increasing the staff and selected to remain lean, however issues had been dire round this time final yr. A deliberate launch in India was stymied by rules, placing the corporate in monetary pressure, a lot so {that a} handful of workers went on strike and claimed they weren’t getting paid. Lyu says he revered their determination, including that the corporate secured extra funding and is on stronger footing. “There are not any points with the corporate’s well being,” he says.
Lyu says the staff determined to not rush into new {hardware}, and as a substitute centered on delivering software program updates to shine the R1 expertise—up to now, Lyu says Rabbit has deployed round 50 updates up to now yr and a half. He factors to a devoted Discord neighborhood of greater than 12,000 customers that present suggestions on what new options so as to add to the R1, like opening up the {hardware}. For instance, you’ll be able to combine OpenClaw or Hermes AI brokers into the R1 as a substitute of utilizing Rabbit’s proprietary fashions.
