Rabbit R1 founder defends ‘unfinished’ AI gadget
When Jesse Lyu unveiled Rabbit’s flagship device, the R1, in January, the response was swift.
The neon-orange, compact AI-powered gadget quickly gained attention for its playful design, selling over 100,000 units in just a few months – a feat that founder Lyu has hailed as “incredible for a first-gen product.”
But many of those who shelled out $199 (£150) for the R1 just as quickly labelled it “half-baked” and “barely reviewable,” with some even going as far as calling it a scam. A glance at Reddit boards and tech forums reveals a flood of complaints about functionality and speed from disgruntled users.
Why would you compare a new technology to something that’s been there for 20 years? Remember the first iPhone; there was no app store.
Most of the criticism is levelled at the device’s promise to be a ‘personal assistant’, moving beyond simple voice commands to handle more complex tasks. “Our focus is on actions, not just language,” Lyu said in an interview with City AM.
While the R1 can understand a user’s description of a task, many users question whether it can actually execute those tasks – like ordering a product on Amazon – without error and if it really goes further than a regular phone.
Lyu, however, believes the naysayers have misunderstood. The first car ran slower than a horse, he said, insisting it’s too soon to compare Rabbit’s R1 to the likes of Apple.
“It’s a new technology. Why would you compare a new technology to something that’s been there for 20 years? Remember the first iPhone; there was no app store,” he said.
“It took Apple three years to introduce the 3GS [iPhone] with the first generation of its App Store and then they blew up,” Lyu added.
He also suggested it is unfair to hold a 17-person startup to the same standards as global tech behemoths, although he now has a team of 60.
One of the most common complaints about the R1 is that it feels unfinished – shipped to market before the software was fully developed. But Lyu brushed this aside, saying: “Go take a walk to Charing Cross or Tottenham Court Road, you have the gigantic Apple poster about the iPhone 16 Pro.
“Which part of Apple Intelligence is ready?” he added, referencing the latest iPhone reviews, many of which complain its most hyped features aren’t yet operational.
He may have a point. Popular American Youtuber and tech reviewer Marques Brownlee said in his recent iPhone 16/16 Pro review: “There’s something just off about billboards and marketing material all highlighting Apple Intelligence, and Tim Cook saying this was the first phone built from the ground up for Apple Intelligence. Yet, this phone does not have Apple Intelligence at all.”
This has been a common trend in tech, Brownlee said, also labelling Apple’s new product as “barely reviewable”, although he added that at least the iPhone is a usable product and the company has promised rapid updates.
It can make mistakes, that’s why we call it a playground.
On the industry trend of shipping hardware with the promise of future software updates, Lyu said: “What people don’t understand is that this generation of AI technology, working with all these models, there is never going to be a chance for any company to say ‘it’s ready’.”
It comes as Rabbit has also just launched its latest innovation: the newest version of Rabbit’s large action model (LAM), called LAM Playground. It’s meant to process words like a large language model does, but then turns them into actions, using apps for you like a human would.
However, during a demonstration, the system stumbled slightly, entering a non-existent word into NYT’s Wordle game before freezing. “There’s a lot of optimisation we’re going to do in the future,” Lyu admitted. “It can make mistakes, that’s why we call it a playground.”
When a more advanced version of the system was asked to send an email to a colleague, it spat out a jumble of letters, misspelled the email that had been typed in, placed the subject line incorrectly and failed to deliver the message.
Lyu still wasn’t discouraged. “We believe we can make it significantly faster”, he said, “but if you push everything too fast then you lose a bit of accuracy.” He still sees the R1 as the next generation of all computing devices.