Rabbit Unveils Android AI Agent Amid Challenges in AI Hardware Market

Rabbit AI's new tool can control your Android phones, but I'm not sure how I feel about letting it control my smartphone

It's currently being tested for Android ahead of a teased release Rabbit is showing off a new AI agent that can automatically carry out tasks for you on Android devices. This might be because Rabbit is is keen to show that it has another trick after Humane's sale to HP renewed focus and questions about the company's Rabbit R1 hardware. Perhaps aware that the hardware approach wasn't selling as well as anticipated, Rabbit has hopped away from hardware to smartphones, where far more people are likely to experiment with AI tools. Rabbit's Android agent doesn't just remake the same features of so many AI chatbots, either. Instead of summarizing Wikipedia pages, Rabbit's agent can coordinate tasks across different apps based on your prompts. You could ask it to write and send dinner party invitations on WhatsApp, pick a meal from a recipe app, or assemble a grocery list based on that recipe. It can also find and save a YouTube music playlist for when you eat and even download a game from the Play Store for after the meal. To really show off Rabbit's AI agent, the demonstration put together by the company showed how the agent could find a business idea you saved in Google Keep and write a revenue plan in Google Docs before sharing it with someone in your contact list. This "doing" AI, as opposed to the "knowing" AI, is usually what people imagine when you describe an AI assistant. But, like any new tech, it's not perfect yet. Rabbit's demo shows that while the agent can pull off these tasks, sometimes it stumbles. Some actions take longer than expected, formatting can get weird, and every now and then, it makes a choice that raises an eyebrow. Rabbit's agent isn't generally available yet. Based on the demonstration, once it is, you'll be able to get the app from the Play Store and link it to Android apps once you give permission. You'll then be able to prompt the AI to complete a task for you, and it will run in the background, which is a nice change from some AI tools that demand you leave the app open while it runs. Rabbit will have to contend with more competition with this move. A growing wave of AI tools offers to take control of your devices and automate your tasks. OpenAI has been developing the app-controlling Operator, and Anthropic is testing a feature called computer use for its Claude AI assistant, which allows an AI to interact with desktop software and the internet. There's also third-party Browser Use, which specializes in navigating the web, filling out forms, and completing requests online. Rabbit isn't imitating any of them precisely, focusing on Android devices and apps. Instead, Rabbit's agent seems geared chiefly toward everyday use compared to OpenAI and Anthropic's emphasis on professional activities. This reimagining of what Rabbit can offer is exactly the sort of reinvention many have suggested could keep Rabbit independent. The standalone R1 didn't have much momentum after it came out and often felt like a limited smartphone with an odd feature set. Now, instead of asking people to buy new hardware, Rabbit is meeting them where they already are: on their phones.

Remember the Rabbit R1? The company just introduced a new Android AI agent.

A new Android AI agent has been unveiled by Rabbit, the company behind the physical Rabbit R1 AI device. Credit: Rabbit.tech Do you remember the AI wearable gadgets that were all the rage last year? Two of the biggest physical products just resurfaced this week after falling out of the limelight following underwhelming releases: Humane and Rabbit. Humane announced it was killing its Ai Pin less than one year after launching it, effectively confirming it was a failure. Rabbit, however, appears to still be kicking. In fact, the company just released a preview video that showed off a demonstration of its new Android AI agent. Rabbit's Android AI agent is able to control an Android device based on the prompts provided by the user. According to Rabbit, with these capabilities its Android agent is able to change app notifications in the device's settings, search YouTube, add ingredients from one app to a list on Google Keep, generate a poem using AI and then sent it in WhatsApp, download a game from Google Play and learn how to play it, and more. The company says this Android AI agent builds on the work that they showed off last year involving its web agent LAM Playground. Watching the video, it's clear that the Android AI agent is very much a work in progress. The AI generally delivers on each user request but with varying degrees of competence. The whole thing is pretty slow though. It's unclear exactly why you would use an AI agent to do these specific tasks mentioned in the video as these are things that can be accomplished by the user basically in the same timeframe it takes to enter the prompt and have the AI deliver it. The odd thing about the whole video is that the company's actual product that it sold to customers, the wearable AI agent known as Rabbit R1, is just sitting unused on a desk. It serves no purpose. The product that's actually available and being sold to customers for $199 is not involved in the performance of any of these tasks being demoed. After seeing Humane crash and burn so quickly, this video shows that Rabbit is still alive and very much working on AI products, so that's a good thing for consumers who have invested in the product already. For now.

Rabbit shows off the AI agent it should have launched with

Wes Davis is a weekend editor who covers the latest in tech and entertainment. He has written news, reviews, and more as a tech journalist since 2020. The Humane AI Pin has collapsed, but Rabbit is still kicking. The company published a blog post and video today showing off a "generalist Android agent," slowly controlling apps on a tablet in much the same way that Rabbit claimed its R1 device would over a year ago. (It couldn't, and can't.) The work builds on LAM Playground, a "generalist web agent" Rabbit launched last year. The engineers don't use the Rabbit R1 at all for the demonstration. Instead, they type their requests into a prompt box on a laptop, which translates them to actions on an Android tablet. They task it with things like finding a YouTube video or locating a whiskey cocktail recipe in a cocktail app, gathering the ingredients, and then adding them to a Google Keep grocery list. At one point, they ask it to download the puzzle game 2048 and figure out how to play it, which it does, albeit slowly. The model generally does the things they ask, sometimes well and sometimes with quirks like sending a poem over WhatsApp one message at a time instead of in a single block. One of the engineers wonders if they should have asked it to use line breaks in their prompt, but they don't go back to try again. Rabbit's AI agent is clearly still a work in progress, as it has been since the R1 launched with almost none of the capabilities that founder and CEO Jesse Lyu presented in January 2024. Rabbit has steadily rolled out updates, like the ability to train its AI agents to complete specific tasks or prompt it to remake its own interface. The examples it presented today are "only the core action loop an Android agent completes," according to Rabbit's blog post. The company promises to share more about its "upcoming cross-platform multi-agent system" in coming weeks.

With the Humane AI Pin now dead, what does the Rabbit R1 need to do to survive?

The excitement around the Humane AI Pin has faded. The company behind the $700 AI device has sold its assets to HP for $116 million, and some of the employees have joined the company. It's a lot of money, but not exactly a coup for Humane, considering the startup had raised over $230 million. HP discontinuing the AI Pin is just the capper. The withering of Humane's fortunes has more significant implications for another AI device, the Rabbit's R1. The chirpy little AI gadget and the startup behind it may end up in a similar place as Humane if it's not careful. There are plenty of reasons why Humane may have failed, though I have a few theories. The biggest reason may have been a lack of features people wanted. There were some intriguing ideas, like the screenless interaction with the AI, the laser projector that looked great in demo videos, and the continual link to the AI assistant. However, the reviews all highlighted slow AI processing, unreliable hardware, and feeling like the $24 monthly subscription fee was too much, especially on top of the $700 purchase price. As AI apps like ChatGPT improved, many may have decided they could get all the features they liked from their smartphone and returned the Humane AI Pin. Rabbit should be taking notes. Right now, the R1 is struggling with some of the exact same issues: a lack of must-have features, performance hiccups, and a growing sense that it's an inessential accessory. Rushing out a prototype too quickly when the hardware is still unstable is a problem, especially when the hardware is the main selling point. A lack of certain features, like messaging and social media support, raises the same questions about its value as Humane faced. I think the R1 can survive if Rabbit makes the right moves. That includes upgrading the R1's Large Action Model (LAM) AI engine. The LAM is supposed to be the magic sauce that makes the R1 superior to Siri or Alexa, but even being just as good won't help it stand out. It needs to be instant and accurate when booking a ride, answering a question, or summarizing emails. Rabbit should throw everything they have into improving LAM, even if that means bringing in outside help and partnering with OpenAI or Anthropic. Rabbit also needs to get deals in place to work with WhatsApp, Slack, and the Android and iOS messaging services at the very least. Even better, they should lean into voice-based messaging and AI-powered summaries, though they'd need to make sure they work at least as well as Apple Intelligence. Ultimately, Rabbit needs to work out its positioning. Right now, it's unclear who the R1 is for. Is it a productivity tool? A digital assistant for busy professionals? A toy for tech enthusiasts? Instead of trying to be everything to everyone, they should focus on a specific audience and go all in. If they position the R1 as an ultra-efficient personal AI secretary, something that can handle voice-to-text notes, summarize meetings, and manage calendars in a smoother way than any smartphone assistant, they might find a real niche. And let's not forget developers. Rabbit should aggressively court third-party developers, giving them the tools to build new integrations and apps for the R1. A strong developer ecosystem could turn the R1 from a novelty into a platform. The Rabbit R1 isn't doomed yet, but it is at a crossroads. It risks becoming another Humane AI Pin, an ambitious but ultimately forgettable attempt at AI hardware. But if Rabbit takes decisive action, it might just pull a, well, rabbit out of its AI hat.