I hate the Rabbit R1. It's yet another sign that standalone AI gadgets, like the Humane AI Pin, are fundamentally useless devices meant to attract hype and VC funding without benefitting users at all. It's like trying to build a skyscraper on quicksand: Today's AI models are great for parlor tricks, but they're ultimately untrustworthy. How do you create a device around that?

The Rabbit R1's big selling point has been its "large action model," or LAM, which can supposedly understand what you say and get things done. But really, that's just marketing speak. At the moment, the R1 can barely do anything as an AI assistant. And the few tasks it can actually accomplish, like placing DoorDash orders, are faster and easier to tackle on your phone. You know, the device we already own that can tap into AI features and fast cellular networking.

Rabbit R1: design and build

I'll admit, the Rabbit R1 looks adorable, but that's mostly down to the design magic of Teenage Engineering, a company that can make a simple tripod look desirable. The R1 is clearly building on the Playdate, another tiny square gadget from Teenage Engineering. Instead of that game handheld’s iconic crank, the R1 has a far less satisfying scroll wheel. Its glossy plastic case also feels a lot cheaper and thicker than the Playdate, almost like what you'd expect from a child's toy.

Alongside the dull 2.9-inch screen, there's a unique 8-megapixel "360 eye" camera, which can rotate either towards you or away from you. It's an interesting way to avoid bundling two separate cameras, so I'll give Rabbit credit for that. But the 360 eye isn't meant for taking photos: Instead, it's all about computer vision. You can ask the R1 to describe what's in front of you, from objects to documents and articles, and wait for an AI-generated summary. While this is something that could be useful for people with visual impairments, those users could do the same with ChatGPT, Microsoft's Copilot or built-in tools on their phones (which also have vastly superior cameras).

Using the Rabbit R1 is an exercise in futility

Beyond its looks, the Rabbit R1 is mostly a failure. Once it’s turned on, you should be able to hit the push to talk button on its side and ask the AI assistant whatever you want: the weather, local traffic or a summary of a recent book. In my testing, though, the R1 would often deliver the weather when I asked for traffic, and sometimes it would hear my request and simply do nothing.

The R1 becomes more frustrating the more you use it: Its scroll wheel is the only way to interact with its interface (even though the display is also a touchscreen), and it's simply awkward to use. There's no rhyme or reason for how long you need to scroll to move between menu options. The mere act of selecting things is a pain, since the confirmation button is on the right side of the R1. That button would be far easier to hit somewhere below the scroll wheel — or better yet, just let me use the damn touchscreen!

Photo by Devindra Hardawar/Engadget

Oddly, the Rabbit's touchscreen does recognize taps whenever you need to enter text like a Wi-Fi network password. But even that process is annoying, since it involves turning the R1 on its side and typing on a laughably tiny keyboard. Honestly, I felt like I was being punked every time I had to use it. (Cue the obligatory, "What is this, a keyboard for ants?")

Third-party apps on the Rabbit R1

The more I used the Rabbit R1, the more I felt like it was purposefully designed to drive me insane. It can play music from Spotify (if you have a paid subscription), but what's the point of doing that with its terrible 2-watt speaker? Are you expected to connect Bluetooth headphones? You can ask the R1 to generate art via Midjourney AI (again, with a paid account), but it often failed to show me the pictures that were created. On the rare occasion they did show up, I couldn't actually do anything with the AI pictures from the R1. I'd have to load up Midjourney's Discord server on my phone or computer to share them around.

Photo by Devindra Hardawar/Engadget

When I asked the R1 to find me an Uber to a local theater, it told me that the Uber service may be slow to load on RabbitOS and isn't available everywhere (uh, thanks?). After 30 seconds of idling, it said the Uber service may be under maintenance, or there may be an issue with my credentials. (I logged out and back into Uber on the "Rabbit Hole" website, which you use to manage the R1, but the error persisted.)

“LAM works by operating the Uber web app on the cloud on your behalf,” Rabbit representative Ryan Fenwick told me over e-mail when I asked why I couldn’t get the Uber service to work. “Uber ultimately decides how and whether it serves users, so depending on factors like the location you’re booking from, your ride history, etc., it may vary from time to time. We’re implementing measures that help to improve the success rate and transparency of ride booking through R1, so over time the experience should improve.”

At least the Rabbit R1 was able to get me a sandwich. I asked it to find some lunch nearby and it spent an entire minute communing with Postmates and its AI cloud — the precise amount of time it would take me to complete a GrubHub order on my phone. The R1 eventually returned with three chaotic choices: Subway, a nearby Henri’s Bakery and a restaurant five miles away I’ve never heard of

Photo by Devindra Hardawar/Engadget

I opted for Henri’s (they do make killer sandwiches), and the R1 showed me a whopping six menu items. Its tiny screen could only hold a picture of the item, its name and the price — you can't tap into it to get a longer description or customize anything. You can only add items to your cart or remove them. I chose two sandwiches and, to my surprise, the R1 completed the order without ever confirming my payment information or delivery address. It was working entirely off of my DoorDash defaults, and thankfully those were up to date.

As soon as the order was placed, my iPhone started lighting up with all sorts of useful information from DoorDash. I received a confirmation from the restaurant, a detailed look at the bill (the R1 apparently added my default 20% tip) and the name of my delivery driver. It took the R1 several minutes before it confirmed the order, and it only occasionally updated me that it was coming closer.

My sandwiches eventually arrived, but I was more struck by the many ways things could have gone wrong. This isn't 1999; I'm no longer impressed by simply being able to order food online like I did from Kozmo.com (RIP). But even back then, I was able to get a full look at menus and customize things. The fact that I could look over at my phone and see the DoorDash app being far more useful made me instantly lose faith in the R1.

There are other things the R1 can do, like recording and summarizing meetings. But that’s also something several apps can do on my phone and computer. The on-demand translation feature seemed to work fine converting English to Spanish and Japanese, but it’s no better than Google Translate or ChatGPT on my phone.

Photo by Devindra Hardawar/Engadget

What’s the point of the Rabbit R1?

All of this leads me to ask: What's the point of the Rabbit R1, really? it certainly can't replace your phone, since it can't make calls or send texts. While you can add a SIM card for always-on connectivity, that just makes it more expensive. It'll still be useless on the go, anyway. Perhaps, you could argue, it's a companion device to help avoid being distracted by your phone. But it's so slow and hard to use that I find my smartphone's notification-filled hellscape far more calming. There's nothing zen at all about having yet another device that you have to buy, charge and carry.

And if you suffer battery life anxiety, you absolutely should stay away from the Rabbit R1. When I first received it, the R1 would burn through its battery while sitting idle, doing absolutely nothing, for eight hours. The first major RabbitOS update helped considerably, but the R1 still can’t last an entire day on a single charge. For a device that has such a tiny screen and offloads its work to the cloud, that’s simply inexcusable.

Photo by Devindra Hardawar/Engadget

I suppose you could argue that the $199 Rabbit R1 is a good deal compared to the $699 Humane AI Pin (which also requires a $24 monthly subscription), but that’s like saying rabbit droppings don’t smell bad compared to dog poop. Technically true! But in the end it’s all still shit. The Humane’s projection screen is at least an interesting twist on mobile UI, and its potentially less cumbersome as a wearable. The Rabbit AI assistant, on the other hand, is basically just a chunkier and dumber phone.

Don’t buy the R1. Even if Rabbit somehow manages to deliver on some of the promises of its LAM – like the ability to train the R1 to handle the variety of tasks – I have no faith that it’ll actually work well. My advice extends to every standalone AI gadget: Just stay away. Your phone is enough.

Epic Games won its antitrust lawsuit against Google in December when a federal jury found that the latter violated US antitrust laws with regards to how it runs the Play Store. A few months later, the gaming developer submitted its list of demands, which if implemented will blow the Play Store wide open. Now, Google has filed an injunction telling the court that no, it will not give Epic what it wants without a fight, because the company's asks "stray far beyond the trial record." 

The remedies Epic had submitted would require the court not just to create a global regulatory regime to set prices for apps, Google wrote in the filing as seen by Engadget, but also to micromanage "a highly complex and dynamic ecosystem" used by billions of consumers and app developers around the world. If you'll recall, Epic wants Google to open up Android to third-party app stores and to make its catalog of apps available to those stores. It also wants restrictions on pre-installed apps to be outlawed and to prohibit any Google activity that incentivizes third-parties. 

Google said that bowing down to all those demands would "effectively prevent [it] from competing," which in turn would negatively affect Android users and developers. Epic's proposals only benefit Epic, Google said in its filing, and will harm other developers by depriving them of control over where their app is distributed. Manufacturers will no longer be able to take advantage of the partnerships Google typically offers, while users have to deal with additional security and privacy risks. 

The company also slammed Epic over the "vagueness" of its proposed injunction, which would require the repeated and ongoing intervention of the courts. Similarly, Epic's demands would apparently require the court to micromanage Google's business. 

"Epic’s demands would harm the privacy, security, and overall experience of consumers, developers, and device manufacturers," Wilson White, Google's Vice President of Government Affairs & Public Policy, told Engadget in a statement. "Not only does their proposal go far beyond the scope of the recent US trial verdict — which we will be challenging — it’s also unnecessary due to the settlement we reached last year with State Attorneys General from every state and multiple territories. We will continue to vigorously defend our right to a sustainable business model that enables us to keep people safe, partner with developers to innovate and grow their businesses, and maintain a thriving Android ecosystem for everyone."

Google said that if Epic truly wants to promote competition rather than create "an unfair, court- supervised advantage for itself," then it would take cues from its settlement with the state officials that previously accused the company of abusing its dominance on Android app distribution. Epic Games CEO Tim Sweeney was, unsurprisingly, unhappy with that settlement, tweeting at the time: "If Google is ending its payments monopoly without imposing a Google Tax on third party transactions, we'll settle and be Google's friend in their new era. But if the settlement merely pays off the other plaintiffs while leaving the Google Tax in place, we'll fight on. Consumers only benefit if antitrust enforcement not only opens up markets, but also restores price competition."

The Rabbit R1 is finally here, and it's yet another useless AI gadget. Sure, at $199 with no monthly fee, it's a lot cheaper than the $699 Humane AI Pin. But the R1 is slow, hard to use, and doesn't actually do much. The much-promised "Large Action Model" mostly powers things you can easily do on your phone. In this episode, Devindra and Engadget's Sam Rutherford chat with CNET's Lisa Eadicicco about the Rabbit R1 and whether AI devices are necessary at all. Just like cameras, the best AI device is the one you always have with you: your smartphone.

Listen below or subscribe on your podcast app of choice. If you've got suggestions or topics you'd like covered on the show, be sure to email us or drop a note in the comments! And be sure to check out our other podcast, Engadget News!

Topics

• Rabbit R1 review: appealing design, underwhelming performance – 0:49

• Tesla lays off Supercharger development team leaving future of the network unclear – 25:28

• FCC fines U.S. wireless carriers $200m for selling customer location data – 30:05

• Razer will refund all Zephyr mask purchases over false N95 filtering claims – 32:52

• Drake deletes track featuring an AI clone of Tupac Shakur’s voice – 35:16

• Working on – 36:23

• Pop culture picks – 40:18

Subscribe!

• iTunes

• Spotify

• Pocket Casts

• Stitcher

• Google Podcasts

Livestream

Credits 

Hosts: Devindra Hardawar and Sam Rutherford
Guest: Lisa Eadicicco from CNET
Producer: Ben Ellman
Music: Dale North and Terrence O'Brien

The Rabbit R1 is finally here, and it's yet another useless AI gadget. Sure, at $199 with no monthly fee, it's a lot cheaper than the $699 Humane AI Pin. But the R1 is slow, hard to use, and doesn't actually do much. The much-promised "Large Action Model" mostly powers things you can easily do on your phone. In this episode, Devindra and Engadget's Sam Rutherford chat with CNET's Lisa Eadicicco about the Rabbit R1 and whether AI devices are necessary at all. Just like cameras, the best AI device is the one you always have with you: your smartphone.

Listen below or subscribe on your podcast app of choice. If you've got suggestions or topics you'd like covered on the show, be sure to email us or drop a note in the comments! And be sure to check out our other podcast, Engadget News!

Topics

• Rabbit R1 review: appealing design, underwhelming performance – 0:49

• Tesla lays off Supercharger development team leaving future of the network unclear – 25:28

• FCC fines U.S. wireless carriers $200m for selling customer location data – 30:05

• Razer will refund all Zephyr mask purchases over false N95 filtering claims – 32:52

• Drake deletes track featuring an AI clone of Tupac Shakur’s voice – 35:16

• Working on – 36:23

• Pop culture picks – 40:18

Subscribe!

• iTunes

• Spotify

• Pocket Casts

• Stitcher

• Google Podcasts

Livestream

Credits 

Hosts: Devindra Hardawar and Sam Rutherford
Guest: Lisa Eadicicco from CNET
Producer: Ben Ellman
Music: Dale North and Terrence O'Brien

Elon Musk has made no secret that he’s not a fan of the block button. Now, X is changing how blocks work on its platform, though it’s (for now) stopping short of Musk’s wish of nuking the feature entirely.

Instead, X is tweaking the visibility of replies in the context of a block. Previously, a user could block someone on X and still reply to their posts. And the person who was blocked wouldn’t be able to view that reply or know the person who had blocked them was engaging with their posts. That’s now being rolled back so that users will be able to see direct replies even if the person posting has blocked them. In a post from X’s engineering team, the company said the change is “part of our ongoing commitment to aligning the block feature with our principles as a public town square.”

The post also hinted that there may be more changes coming to blocks, which is one of the most widely used safety features on the service. “Our goal is to allow users to control their experience while maintaining the public visibility of posts,” the company said. That would seem to align with previous comments from Musk, who has said on multiple occasions that a public-facing “block” feature “makes no sense” and that “block is going to be deleted as a ‘feature,’ except for DMs.”

For now, the changes to blocking aren’t that drastic. In fact, they may even be welcomed by some users who want more visibility into what people are saying about them on the platform. But the fact that X is starting to change the visibility of previously blocked content could also be a sign of more significant changes to come.

Microsoft says the April security updates for Windows may break your VPN. (Oops!) “Windows devices might face VPN connection failures after installing the April 2024 security update (KB5036893) or the April 2024 non-security preview update,” the company wrote in a status update. It’s working on a fix.

Bleeping Computer first reported the issue, which affects Windows 11, Windows 10 and Windows Server 2008 and later. User reports on Reddit are mixed, with some commenters saying their VPNs still work after installing the update and others claiming their encrypted connections were indeed borked.

“We are working on a resolution and will provide an update in an upcoming release,” Microsoft wrote.

There’s no proper fix until Microsoft pushes a patched update. However, you can work around the issue by uninstalling all the security updates. In an unfortunate bit of timing for CEO Satya Nadella, he said last week that he wants Microsoft to put “security above else.” I can’t imagine making customers (temporarily) choose between going without a VPN and losing the latest protection is what he had in mind.

At least one Redditor claims that uninstalling and reinstalling their VPN app fixed the problem for them, so it may be worth trying that before moving on to more drastic measures.

If you decide to uninstall the security updates, Microsoft tells you how. “To remove the LCU after installing the combined SSU and LCU package, use the DISM/Remove-Package command line option with the LCU package name as the argument,” the company wrote in its patch notes. “You can find the package name by using this command: DISM /online /get-packages.”

Audible is testing a cheaper subscription tier in Australia that sounds like an answer to Spotify’s audiobook push. Like the more expensive plan, the new Standard membership gives you one free title per month. But, similar to gaming services like PlayStation Plus, you lose access to the books you claimed under the plan if you cancel.

As reported by Bloomberg, the Audible Standard plan costs AUD 8.99 (US$5.90) per month, significantly lower than the AUD 16.45 monthly cost of the Premium Plus plan in Australia ($14.97 in the US). Despite the lower cost, Audible Standard subscribers down under can still claim one audiobook per month to add to their library, which they can download and listen to online or off.

But the catch is, unlike Audible Premium Plus, Standard members’ free audiobook credits don’t roll over to the next month if they don’t use them. (So, if you forget to claim a book in May, you’ll still only have one credit to use in June.) In addition, the audiobooks you chose on the Standard plan will have a lock icon next to them after you cancel. To listen to them after canceling, you’ll have to buy them at full price or resubscribe.

Another difference between the plans is Audible Standard members won’t get full podcast access. “Audible Standard members can listen to many podcasts for free,” an Audible FAQ about the plan reads. “Some Audible Original podcasts aren’t available with Audible Standard membership though.”

Spotify launched an audiobook feature last year that challenged Audible’s established model. Spotify Premium subscribers in the US can stream 15 hours of books monthly through their ad-free music plan. The service also offers an audiobooks-only tier (with the same 15-hour cap) for $10 monthly. If the allotted time isn’t enough, Spotify users can buy a 10-hour top-up. A single audiobook often lasts around seven to 11 hours.

Microsoft is celebrating World Password Day (IT folks deserve holidays, too!) by helping to kill them. The company has finally rolled out consumer passkey support for Microsoft accounts, nearly two years after Apple and Google.

Once you set it up, the passkey lets you sign into your Microsoft account using your face, fingerprint or device PIN. It works not only on Windows but also on Apple and Google’s mobile and desktop platforms.

Passkeys are an easier and more secure way to access your account. They use what’s called a cryptographic key pair to ensure only you can get in. One half of the pair is stored on your local device, only accessible via your secure local login. The other part stays on the app or website. Requiring both of them to sign in acts as a deterrent for things like password leaks and phishing attacks.

In addition to Apple, Google and now Microsoft, companies adopting passkeys include Amazon, 1Password, Dashlane, Docusign, eBay, PayPal and WhatsApp (among others). Google said on Thursday that its passkeys have already been used a billion times.

Microsoft’s passkey support works starting today on the company’s desktop apps and websites, including Microsoft 365 (Office) and its Copilot AI assistant. The Windows maker says passkey support for its mobile apps “will follow in the coming weeks.”

You can get started by signing into your Microsoft account here and following the instructions.

Bose may be best known for its noise-canceling headphones, but the company makes solid portable Bluetooth speakers too. In fact, the company's SoundLink Flex made our best Bluetooth speakers list as a great option among contenders in its price range. Today, the company is adding to the SoundLink lineup with its largest portable Bluetooth speaker yet: the SoundLink Max ($399). While the overall design is similar to previous Bose devices, this model packs bigger sound and longer battery life into that expanded frame. 

Inside the SoundLink Max, three transducers and two passive radiators power "a spacious stereo experience" that includes bass performance that sounds like an even larger speaker, according to Bose. The company says this portable unit employs tech typically used in its soundbars and pairs that with digital signal processing to reduce distortion for "full, natural sound" across genres. Bose is promising that you'll be able to hear every aspect of a song clearly, no matter the musical style and no matter where the speaker is located. You'll also be able to adjust lows, mids and highs via the Bose app if the stock tuning doesn't suit your preferences. 

Bose

Bose opted for a powder-coated, silicone-wrapped steel enclosure for the SoundLink Max, which the company says offers a more refined look. The speaker is also IP67 rated, so dust, water, rust and dropping it shouldn't be an issue. This all makes the Max well-suited for outdoor use, and when you do take it on the go, you won't have to worry about recharging often. Bose says the SoundLink Max will last up to 20 hours, plus it can juice up your phone via a USB-C cable if needed. A removable rope handle will assist with transport, but Bose also makes a carrying strap if you prefer over-the-shoulder hauling. 

The SoundLink Max is equipped with Snapdragon Sound, which offers more consistent connectivity with recent Android devices, and aptX Adaptive that provides improved audio quality over Bluetooth. The speaker also supports Google Fast Pair and Bluetooth 5.3. 

Pre-orders for the SoundLink Max start today from Bose, and the speaker is schedule to ship on May 16. In addition to being the company's largest portable Bluetooth unit, it also ties the Bose Portable Smart Speaker for being the most expensive at $399. If you're looking for something smaller, the SoundLink Micro ($99), SoundLink Flex ($119) and SoundLink Mini II ($149) are also available from Bose. 

As part of World Password Day (yes, that's a thing), Google is hyping up its security achievements and sharing updates on its latest endeavors. The company revealed passkey adoption metrics for the first time and said they have been used more than one billion times by 400 million Google accounts. "Since launching, passkeys have proven to be faster than passwords, since they only require users to simply unlock their device using a fingerprint, face scan or pin to log in," Google wrote.

The company launched broad support for passkeys in 2022 and rolled them out across its services a year ago. Over the past 12 months, the technology has been adopted by Amazon, 1Password, Dashlane, Docusign and others, joining companies like eBay, PayPal and WhatsApp. Google boasted that the tech helped Kayak users sign in 50 percent faster and said Dashlane has seen a 70 percent increase in conversion with passkeys. 

Google will soon be expanding passkeys to users at the highest risk of targeted attacks as part of its Advanced Protection Program (APP). That offering is aimed at individuals including campaign workers and candidates, journalists, human rights workers and others, according to the company. 

"APP enrollment traditionally required the use of hardware security keys as a second factor, but users will soon have the option to enroll with any passkey in addition to using their hardware security keys," Google wrote. "This expanded passkey support will help reduce the barrier of entry to APP while still providing phishing resistant authentication... [and] is coming during a critical election year." 

It's also expanding Cross-Account Protection to safeguard users on multiple platforms. That system lets Google share security notifications about suspicious event with non-Google apps and services. "This is a critical benefit since cybercriminals often use an initial entry point as a foothold to gain access to more of your information." 

Google suggests creating a passkey for your account to benefit from the new protections. In the meantime, practice good password hygiene by using long passwords with a mix of characters, numbers and symbols, applying two-factor authentication (2FA), never recycling passwords and more. According to HIPAA, attackers can crack a simple 8-number password in just 37 seconds, but it takes 19 quadrillion years to break an 18-digit cypher with a mix of numbers, upper and lowercase letters and symbols.