Google’s Circle to Search is getting a new feature that will allow you to continuously translate text while scrolling through social media posts or even when switching to different apps.
Announced on Thursday, the feature will start its rollout this week to select Samsung Galaxy devices, but will more than likely expand to more devices, with the Pixel 10 lineup a natural next stop.
While you could already translate text with Circle to Search, the new update will make the experience more seamless. Previously, if you had multiple items that you wanted to translate that were either in different apps or required scrolling to get to, you’d need to redo the same translation process for each one of them. The latest update rids you of having to do those extra steps.
The updates to Circle to Search make continuous translation faster and less finicky.
Google
Google’s AI-powered Circle to Search launched in early 2024 on the Pixel 8 and Galaxy S24 smartphones and felt like an easier-to-use Google Lens. It’s handy for shopping when you find something you want more information about while watching a video or scrolling through social media.
It’s essentially another way to perform a Google search without leaving the app you’re currently using. Circle to Search has received new features and expanded to more devices since it debuted.
Don’t miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source on Chrome.
How to translate and scroll with Circle to Search
When it rolls out, using the new feature is easy and requires only an extra tap. Here’s how:
Activate Circle to Search by holding down the gesture bar or home button on your phone.
Tap the Translate icon.
Tap Scroll and translate.
That’s it. Once you tap into the continuous translation mode, you’ll be able to get to the information you want faster.
For more, don’t miss Google’s Gemini expansion into its smart home devices.
Nvidia has announced that its Geforce NOW streaming platform will be getting Blackwell RTX update on September 10th, offering Geforce RTX 5080-class power streaming from the cloud, over 2,200 new Install-to-Play titles and more, as well as 17 fresh games this month and Battlefield 6 in October.
The Nvidia Geforce NOW Blackwell update, which was detailed during Gamescom 2025, includes both CPU and GPU upgrades, both adding more performance as well as DLSS 4 Multi-Frame Generation support and pushing streaming up to 5K resolution and 120 fps. It also adds support for 360 fps at 1080p resolution with Nvidia Reflex, Cinematic-Quality Streaming and Install-to-Play, and more.
Most importantly, the GeForce NOW Ultimate remains available for $19.99 per month or $99.99 for six months, and includes access to RTX 5080 servers. Performance memberships remain available at $9.99 per month or $49.99 for six months. Daily, monthly and annual membership will be available starting at launch. Annual memberships will be available for $199.99 and $99.99, respectively.
When it comes to new games, Geforce NOW will be getting Hell is Us, Metal Eden, Hollow Knight Sliksong, Cataclismo, and Cronos: The New Dawn.
The full list also includes:
Firefighting Simulator: Ignite (New release on Steam, Sept. 9)
Borderlands 4 (New release on Steam and Epic Games Store, Sept. 11)
Fata Deum – The God Sim (New release on Steam, Sept. 15)
Dying Light: The Beast (New release on Steam, Sept. 19)
Jump Space (New release on Steam, Sept. 19)
Warborne Above Ashes (New release on Steam, Sept. 19)
ENDLESS Legend 2 (New release on Steam, Sept. 22)
Baby Steps (New release on Steam, Sept. 23)
Aztecs: The Last Sun (New release on Steam, Sept. 23)
Entries are now open for Interop 2026, a yearly effort where anyone, from users to web developers, can submit their best ideas to improve cross-browser interoperability. Here’s how to participate.
Interop started in 2022, when Apple, Bocoup, Google, Italia, Microsoft, and Mozilla came together “to commit to improve interoperability in 15 key areas that will have the most impact on web developer experience.”
The project, which stemmed from Google’s Compat initiative, gained enough traction to become an annual event.
As the group explains it:
“We’re looking for proposals that address real-world interoperability challenges. If you are struggling to get part of your project to work in every browser at the same time, even while coding with proper progressive enhancement techniques, then you are running into an interoperability challenge.
A lot of interoperability problems have already been addressed in Interop 2025, 2024, 2023, 2022, and 2021. Many long-standing bugs have been fixed. So, what else needs to be tackled? What is truly getting in your way? We want to hear from you, the web developer.”
Apple stresses that it is important that any submitted idea is built on top of a mature web standard from the likes of W3C, and TC39, otherwise “it’s too early for it to be a Focus Area for Interop 2026.”
The company notes that testability is also an essential criteria, so it is important to keep these two things in mind.
How to submit an idea
The Interop group has published an extensive documentation on Github, where they also flesh out the criteria outlined above.
But in a nutshell, as explained on Apple’s WebKit blog:
The best submissions are typically:
Specific: Identify a specific interoperability issue. One feature (or narrow group of related features) per proposal. Think font-size-adjust, not “typography”.
Impactful: Provide a clear description of the interoperability problem that needs addressing. Use cases from your own experience are especially helpful.
Valuable: Explain how web developers and/or users will benefit. Why is this more valuable than other options? Is there evidence this is a common need (perhaps survey results)?
Stable: Link to the stable web standard that defines the technology you want improved. (Find it on the MDN web doc for that technology. Like here.)
Tested: Look up and link to WPT test coverage. If more tests are needed, how can they be created this fall?
Apple also says that aside from reading proposals submitted in previous years, participants can either file new issues on Github, or comment on existing proposals “to help further make the case.”
Proposal submissions opened today, September 4, and will run through September 24.
Accessory deals on Amazon
FTC: We use income earning auto affiliate links. More.
As modern military and aerospace operations increasingly depend on seamless communication and real-time intelligence, fiber optics are enabling faster, more secure, and more reliable data exchange. In contrast with traditional copper cables, which are heavier and bandwidth-limited, fiber optics offer lightweight, high-performance connectivity for a range of applications, including remote bases, advanced uncrewed aerial systems (UASs), and wearable tech. Fiber is transforming how personnel, command centers, and autonomous systems stay connected, even in the most demanding scenarios.
The world is rapidly becoming more technologically interconnected, and the military is no exception: Modern military operations are increasingly dependent on having a seamless, secure exchange of information to connect personnel, autonomous systems, command centers, and intelligence networks in real time.
Meeting the needs of communication networks across military bases and other operational areas requires an advanced technological infrastructure – one that has gradually become reliant on next-generation fiber optic cables. The use of fiber optics continues to grow swiftly across all sectors: According to Global Market Insights, the aerospace and defense fiber optics market, valued at $6 billion in 2024, is projected to be worth $15.8 billion by the end of 2034. (Figure 1.)
[Figure 1 ǀ A market forecast shows that the defense and aerospace fiber-optics market continues to grow rapidly, with the expectation that the market will more than double in size over the next decade (2025-2034).]
With high-bandwidth applications like radar, electronic warfare (EW), unmanned systems, and space-based platforms requiring the ability to transmit ever-greater volumes of data across increasingly complex platforms, fiber optics are constantly evolving to meet that challenge.
Fiber versus copper
In decades past, military installations used only copper cable architectures, which tend to be less expensive than fiber optics. The method by which fiber optics transmit data – converting electrical signals into light, sending them along hair-thin strands of glass, then converting them back into electrical signals at the other end – enables signals to travel further than traditional copper systems, with greater bandwidth and less signal degradation.
Fiber’s ability to transmit data over long distances makes it a smart option for military bases, ships, and aircraft, which are often located in isolated or extreme environments where infrastructure is limited or signal amplification is difficult. Fiber is also markedly more lightweight than copper cable, an important distinction for drones, satellites, wearable gear, and mobile command units, where saving weight in wiring can allow for easier transport or create room for other onboard features. (Figure 2.)
Also, crucially, the use of fiber increases the security of classified or mission-critical data communications, since it’s significantly more difficult to tap into than copper. In addition, fiber’s resistance to electromagnetic interference (EMI) makes it a logical choice for radar, shipboard systems, and EW systems because it prevents disruptions both natural (lightning strikes) and human-made (electronic jamming).
[Figure 2 ǀ A fiber versus copper table shows that, especially for military uses where large amounts of data need to be securely transmitted over long distances, fiber optics conveys many benefits over RF cables.]
Replacing an aging infrastructure
While many military installations and networks have already upgraded from older copper cable architectures to fiber optics, an aging infrastructure remains in place in many areas.
Even the fiber-optic cable currently in place at military facilities falls short of the capabilities offered by newer generations of multimode fiber. Today’s fiber is capable of higher speeds and greater data capacity, which are needed to support artificial intelligence (AI)-driven tools, real-time data integration, secure communications and other next-generation military technologies.
The still-common OM1 fiber optic cable was introduced in 1989, and at the time set a new standard for multimode fiber, which is designed to carry multiple light signals simultaneously through its relatively large (62.5 micrometer) core. Over short distances, it can support data speeds of as fast as 10 Gb/sec.
In contrast, the newest generation of multimode fiber optics, OM5, is engineered for high-speed data transmission over multiple wavelengths, enabling a minimum of 28 Gb/sec per channel and as high as 100 Gb/sec over 150 meters (492 feet), making it ideal for modern, data-heavy military or enterprise networks. Its core is also comparatively smaller than OM1 fiber, at 50 micrometers. OM5’s support for wavelength-division multiplexing, which divides multiple streams of information into different wavelengths of laser light, means that a massive amount of data can be sent across military systems with extremely low latency. (Figure 3.)
[Figure 3 ǀ A photo shows OM5 (lime, on right), the most recent generation of multimode fiber optics, which – while it has a slightly smaller diameter than its 1980s-era predecessor OM1 (orange, on left) – is capable of significantly faster data transmission.]
Advancing past obstacles
Over decades of use, fiber optics have advanced to overcome some of their traditional challenges in defense applications. While fiber’s glass core can make it more susceptible to breakage than copper, and connections can be disrupted more easily in high-vibration environments like helicopters or ship engine rooms, today’s fiber cables are well-protected enough to withstand rugged conditions and can be easily spliced or reterminated if damaged.
Aligning two fiber-optic fibers requires extremely high precision to minimize signal loss and preserve the integrity of the laser light. However, advances in connector and termination technology have vastly improved alignment accuracy and reliability, making it easier to maintain high-speed data transmission.
Fiber optics are also susceptible to contamination from particles as small as a speck of dust, which can degrade performance. However, with proper handling, regular inspection, and cleaning, these risks can be effectively minimized, thereby ensuring reliable, high-quality signal transmission even in demanding environments.
Innovations in fiber optics
Continued advancement in fiber optics is positioning the technology for even broader deployment across military and aerospace systems.
New polishing techniques and connector designs are helping to minimize signal loss, which is essential in applications like radar and EW where timing and fidelity are critical. While interconnects have historically been a weak link, these innovations are enabling ultra-low-loss fiber assemblies that can meet even the most stringent performance requirements. This advance will be especially relevant for naval systems, where fiber often needs to span considerable distances across a ship or link remote sensors with minimal latency.
Bend-insensitive fiber (BIF), which addresses fiber’s historical sensitivity to tight turns, has gone from being a specialized product to an industry standard for multimode fiber. BIF can enable a much tighter bend radius when routing in cramped areas, which makes it helpful for use in aircraft avionics bays and on naval vessels. It’s resilient enough that even tight U-turns or loops won’t result in performance loss.
Similarly, branched-fiber configurations continue to evolve as fiber-optic technology advances. These configurations enable more flexible routing and branching of optical fibers, supporting scala-ble and adaptable network architectures. This flexibility helps with system expansion and efficiently uses available space, benefits that are valuable in demanding military and aerospace environments where compactness and reliability are critical.
The next likely innovation in optical fiber branching will involve technologies that enable branching and merging between different types of optical fibers – such as those with varying core diameters or modal properties – while minimizing signal loss and avoiding communication interruptions. These advancements will significantly broaden the range of fiber types that can be interconnected, compared to current limitations.
Meeting the supply challenges of the future
Despite these advances, adoption in the defense world remains gradual, a principal reason being the complexity of military procurement processes, which can involve years of documentation, testing, and approvals.
When upgrades do make it to completion, the benefits can be considerable. One recent example is the ongoing 2025 “Fiber Deep” project on Joint Base Pearl Harbor-Hickam, Hawaii, a base-wide fiber-optic installation undertaking that is expected to save the base as much as $10 million in reduced upkeep and repair costs while increasing network resilience and protecting against cyberthreats.
Military and aerospace connectivity increasingly demands not only high data-transmission speeds and low latency, but also vigorous resilience against interference and unauthorized access. The next generation of fiber optics is capable of delivering all these benefits and more, helping to future-proof military and aerospace networks in order to meet tomorrow’s challenges.
Diana Nottingham is a fiber-optics product line manager for Infinite Electronics, a global provider of connectivity solutions. Diana has more than 15 years of experience in product management and marketing, with a focus on creating and delivering innovation for the optoelectronic and interconnect industry. Infinite Electronics is the parent company for nearly 20 brands, including Integra Optics, which offers reliable transceivers and fiber optic components; and Transtector and Polyphaser, which offer high-performance AC, DC, data, and RF surge protection and connectivity products. Readers may reach the author at [email protected].
Huawei’s second-generation trifold phone is here. The Mate XTs is now available, just shy of a year since Huawei’s first trifold phone was released in September 2024.
The trifold phone isn’t much different from the Mate XT, but there are two big upgrades that take the spotlight. The Mate XTs will let you use a stylus on the phone’s folding screen, and comes in purple, red, black and white. It runs on HarmonyOS 5.1, which Huawei calls a “PC-level” experience, and has a new M-Pen 3 stylus.
“In the free multi-window mode, application windows can be stacked, resized and quickly switched allowing PC-like efficiency at any time,” Huawei said in a press release about HarmonyOS 5.1.
The Huawei Mate XTs.
Huawei
The Mate XTs also reportedly has an upgraded chip, the Kirin 9020. There’s also reportedly a 40MP camera.
The Mate XTs is now available for preorder in China and will launch on Sept. 12. Its starting price is 17,999 yuan (around $2,520), but it could cost more if you want more memory.
But there could be some similar options on the horizon if you’re looking for a foldable phone — Huawei’s announcement comes just weeks before Samsung is expected to launch its Galaxy Z TriFold on Sept. 29. There are also rumors that Apple could launch a foldable phone as a part of its iPhone 18 lineup in 2026.
Stardew Valley creator Eric “ConcernedApe” Barone has a cameo role in Team Cherry’s new mega hit indie game, Hollow Knight: Silksong.
The sequel to Team Cherry’s original bleak and beloved metroidvania game finally launched across PC and consoles earlier today, bringing an end to more than six painful years of anticipation. As what is sure to be millions of fans swarm to see if the wait was worth it, The Verge has pointed out that the Hollow Knight: Silksong credits tease a special cameo from one of the most notable indie developers of the last decade.
Those who navigate to the Hollow Knight sequel’s credits – which can be found in the “extras” option on the main menu – will notice that Barone is featured among the shortlist of names included under “Additional Character Voices.” It means that, if you know where to look, you’ll be able to hear the mastermind behind farming sim sensation Stardew Valley among those lending their voice talents to the land of Pharloom and all of its insect inhabitants.
Hollow Knight: Silksong Credits Feature Stardew Valley Creator Eric “ConcernedApe” Barone. Image captured in Hollow Knight: Silksong by IGN.
Unfortunately, it’s a bit unclear how exactly Barone is featured in Hollow Knight: Silksong. ConcernedApe’s head of operations and business development, Cole Medeiros, confirmed with The Verge that, while the Stardew Valley creator does indeed play a role in the Hollow Knight follow-up, they aren’t ready to reveal who he plays. Medeiros added that Barone would “rather not say which character(s) so as not to spoil any surprises for anyone.”
Although hundreds of thousands of players have already logged in to Steam to play Hollow Knight: Silksong on a Thursday morning, it might take some time to find how Barone is featured. Team Cherry made the metroidvania sequel available for players today, September 4, 2025, for $19.99 for PC, Nintendo Switch, Nintendo Switch 2, PlayStation 4, PlayStation 5, Xbox One, and Xbox Series X | S, and it immediately helped cause digital storefronts like Steam and the Nintendo eShop to crash. As fans continue to pour in, they’ll be met with a game that appears to be substantially bigger than the already quite large original Hollow Knight.
As fans continue to turn over every bug boss and creepy enemy in search of Barone, you can see how players can take advantage of enhanced features in the Switch 2 version of Hollow Knight: Silksong. You can also see why the announcement of Hollow Knight: Silksong’s release date had a few other developers ready to change their own schedules.
Michael Cripe is a freelance writer with IGN. He’s best known for his work at sites like The Pitch, The Escapist, and OnlySP. Be sure to give him a follow on Bluesky (@mikecripe.bsky.social) and Twitter (@MikeCripe).
If you thought AI was already almost indispensable, watch out, because now you can wear an AI gadget designed to record everything you say and hear, and distill it down to a handy summary. These devices promise to give you perfect recall of every conversation. Record every meeting. Settle arguments. Make a to-do list plucked out of your conversations, seemingly like magic. That’s what the best AI wearables like the Bee Pioneer Edition and Plaud NotePin promise, anyway.
AI wearables like Bee, Plaud NotePin, Omi and Limitless promise to streamline your life by recording and summarizing your day.
Illustration: Forbes / Photo: Dave Johnson for Forbes
I tested all of the AI wearables you can buy right now—Bee, Plaud Pin, Omi and Limitless Pendant—for a full month. When you think of these gadgets, some people might imagine an episode of Black Mirror. But the reality was that these gadgets proved to be surprisingly helpful. What’s it like to wear a device that takes detailed notes on your every waking moment? Read on to see what I thought of each of these wearable gadgets.
AI Wearables, At A Glance:
Let’s not bury the lede—here are the best reasons to buy each of these AI pendants:
Bee Pioneer Edition: A great way to get your feet wet in the world of AI wearables; works far better than you have any right to expect for $50, and there’s no monthly subscription fee.
Plaud NotePin: Superb summaries, transcriptions and mind maps, though you’ll need to turn it on and off throughout the day. Even so, it’s my favorite for the workplace.
Limitless Pendant: With short, bulleted summaries that expand into the actual transcription for details, Limitless is probably too concise and not helpful enough for some people.
Omi: Good summaries with a strong emphasis on its app ecosystem and future extendibility. This gadget has a ton of potential, though I do worry a little about its data privacy.
Read on to see what I thought about each one of these AI pendants.
Dimensions: Pill-shaped, 1.8 x 0.8 inches | Charging: USB-C | Data and Security: Not HIPAA compliant, raw audio deleted immediately | Wear style: Clip on, wristband | Recording mode: All day/lifelogging | Battery life: Seven days
Pros:
Elegant UI that shows to-dos and summarizes your day as well as individual meetings and conversations
Ridiculously inexpensive with no ongoing subscription
Always on (unless you mute it)
A full week of battery life
Cons:
You have to approve every to-do and fact about yourself; fall behind and it becomes unmanageable
Frequently hallucinates or misunderstands the details of conversations and meetings
Right now, iPhone only (Android reportedly coming soon)
You can wear the Bee as a pendant—it comes with a clip you can pin to your clothing—or around your wrist like a fitness band. Aside from a mute button, there are no controls on the Bee itself; the Bee just sits in the background, listening and recording. (Unlike the other AI pendants, Bee lights up when muted; all the others display a recording light.)
Like the other AI wearables, all the action happens in the mobile app. Right now, it’s only compatible with iPhone, though an Android version is reportedly coming soon. The app displays daily summaries and suggests to-dos based on what it has heard. You can also use the chat mode to ask Bee questions, like “what day did we agree to go see the new Superman movie?” or “what tasks did my team want to give to the new hire?” That’s an important feature; Bee knows a lot more about your day than you’ll find in the summary, so if you want to know something specific, just ask.
Bee writes almost poetic summaries of your day.
Dave Johnson For Forbes
Among all the wearables, Bee tries the hardest to make you the hero of your personal story. Bee paints a picture of your day with an effusive headline (one of mine: “A day of home, dogs, cars, media, and shared moments with loved ones, punctuated by humor and affection”) and then waxes poetically about your day in the overview. For example: “Your day was a vibrant mix of personal routines, dog care, and a deep engagement with a variety of people whom you work with.” You get “key takeaways”—a bulleted list of the most important events—“atmosphere,” which describes the emotional tone of the day, and summaries of each major meeting and event throughout the day, arranged chronologically with time stamps. The transcript tab offers snippets of dialog throughout the day, organized by topic.
But Bee can be more trouble than it’s worth. It hordes suggested to-dos that you need to manually approve one-by-one to move to the actual to-do tab, and I routinely get dozens of suggestions each day—many of which are teased out of entertainment content that Bee has mistaken for things I said. If you miss a day or two, things spiral out of control and you may have hundreds of suggested to-dos to wade through, enough to make me want to declare AI bankruptcy and toss Bee into a drawer forever.
Right now, Bee is only compatible with iPhone, though an Android version is reportedly coming soon.
Dave Johnson For Forbes
Likewise, it collects facts about me, which I have to review as well. Not only are many of these facts also mistakenly pulled from TV shows, podcasts, and other people I’ve talked to (“Dave is a criminal lawyer,” “Dave has previously worked at a rehabilitation center,” “Dave’s full name is David Elfman,”), but it offers contradictory facts from one day to the next, suggesting it isn’t internalizing its own data—and it’s entirely unclear what purpose this fact collection is supposed to serve, anyway. Bee doesn’t seem to use any of these facts to improve the way it processes information about me, which makes the whole exercise of reviewing these facts seem pointless. On the plus side, for Severance fans, it feels like Bee is telling you facts about your outie.
Speaking of AI hallucinations, Bee isn’t good at knowing who is speaking at any given moment or outright thinks additional people were present for conversations. The good news is that while this is often amusing or perplexing, it generally doesn’t affect the usefulness of the summaries it provides.
Bottom line: Bee is remarkable. It’s roughly as good as the competition for a fraction of the price, and now that Omi has moved to a subscription model, it’s the only pendant that doesn’t have an ongoing fee to pay each month. More than once I’ve used it to win a bet by proving what I said days earlier. It’s a somewhat reliable note-taker in meetings (even though it doesn’t know who’s speaking). I listen to a lot of podcasts, and sometimes it knows I’m listening to content, sometimes it doesn’t. But all this is likely to improve with software updates, so for a first-generation device with a fire-sale price, color me thoroughly impressed.
Dimensions: Button shaped, 1.9 x 0.7 inches | Charging: Proprietary dock | Data and Security: HIPAA compliant, end-to-end data encryption | Wear style: Magnetic clip, optional wristband or lanyard | Recording mode: Per meeting | Battery life: 3-4 days, depending on how many meetings you record
Pros:
Uses ChatGPT and Claude for excellent, accurate and multi-lingual transcriptions
Delivers the most immediately readable and useful summaries for each meeting and conversation
Preserves original recordings as well as transcripts and summaries
You can choose from summary templates for different professionals and applications
Cons:
Most expensive wearable and requires an ongoing subscription
No overall to-do tracking like some AI wearables, though each meeting lists action items
You need to activate it for specific conversations—isn’t always on
Need to wait for recordings to transfer to your phone before summaries are available
The Plaud NotePin is different than the other wearables in this roundup in two important ways. First, it’s not designed to be an always-on, life-logging device, capturing everything it hears all day long. Instead, you press the NotePin until it vibrates (and a small recording light comes on) to start capturing a specific meeting or conversation, and then press again to stop when the meeting is over.
adsf
Dave Johnson for Forbes
Secondly, while other gadgets emphasize their summaries, Plaud makes it super easy to access the original audio recordings as well as complete, aesthetically formatted transcripts in addition to excellent summaries—plus a visual tree-like mind map that organizes the topics from your conversation in a smart and visually engaging way. In fact, Plaud should be applauded for the elegance and usability of its app. And of course, you can use the Ask AI chat window to extract and synthesize information from the conversation as well.
The NotePin is clearly optimized for professionals who need to capture details of interactions with co-workers or clients in vivid detail. Plaud includes an ecosystem of summary templates, which you can choose depending on the kind of meeting or conversation you need to transcribe. The default summary template is excellent and was all I needed for my day-to-day needs, but you can also choose templates created by a Plaud developer community for class lectures, job interviews, construction project meetings and medical sessions (like psychotherapy notes and and medical consultations).
The Plaud NotePin’s summary is excellent but you also get full transcripts and original audio as well.
Dave Johnson For Forbes
Out of the box, Plaud doesn’t automatically transcribe and summarize its recordings; I found that I had to select recently uploaded conversations and ask the app to create a summary. This is because Plaud ties its device to a subscription service—if you don’t have the unlimited plan ($240 per year) then you may need to be selective about which conversations you transcribe and summarize. You can buy 1,200 minutes per month for $99.
If you do have the unlimited plan—or you’re not otherwise worried about running out of minutes—then AutoFlow is a cool feature. AutoFlow is automation you can use to automatically perform certain tasks. I was able to tell NotePin to automatically transcribe and summarize every conversation, for example, and that’s just for starters. I found you could transcribe only conversations with certain keywords, apply specific summary templates based on the kind of conversation, and optionally send emails when the transcription is complete.
In many ways, Plaud seems like the most mature of the AI wearables. While it isn’t intended for always-on recording—something all the other gadgets are designed to do—its polished interface and community ecosystem of summary templates and potential integrations suggests it’s a formidable product.
Dimensions: Button shaped, 1.2 x 0.5 inches | Charging: USB-C | Data and Security: HIPAA compliant, private and secure data | Wear style: Magnetic clip | Recording mode: All day/lifelogging | Battery life: Roughly two days
Pros:
Bulleted summaries accompanied by transcriptions good for encapsulating meetings
Learns voices to distinguish speakers
Subscription tiers include free (20 hours/month) service
Always on (unless you mute it)
Cons:
You don’t get thorough summaries—you need to crawl through transcriptions for detailed notes about meetings
Hyper-focused on summarizing meetings; you don’t get to-dos or broad summaries of the day
Limitless looks like what you get if you connect a pair of Apple Airtags with a rubber hinge. The two discs stick together magnetically, so you can clip the pendant over clothes—but if you prefer, you can slip it onto a lanyard and wear it around your neck. That’s my preference, because the magnet isn’t especially powerful and more than once it slipped off my shirt. It has a side button you can use to start and stop recording.
Like most of the products in this roundup, Limitless is designed to listen in the background all day long. Throughout the day, it uploads your conversations to the phone and groups your summaries with descriptive titles and timestamps for easy reference.
Limitless has very succinct summaries which are not as helpful as some other AI wearables.
Dave Johnson for Forbes
Limitless organizes your conversations a little differently than other AI pendants: Within each conversation, Limitless creates a set of nested bullet points, breaking events into chunks. For example, Limitless divided an hour-long lunch chat into short 15-minute segments and entitled one, “A wide-ranging conversation about electric cars, including selling a Tesla, renting a Nissan, and the politics of tax credits.” Within that summary are a half dozen bullets summarizing part of the conversation, like: “Recounting a recent attempt to sell a Tesla” and “Frustration with a rented Nissan key fob and tire pressure.” Expand each of those bullets, and you get the actual transcribed conversations, divided by timestamp and speaker. It’s a no-nonsense, utilitarian way to encapsulate your day, but it’s not as fun to read as Bee or as elegantly organized as Plaud. And the info isn’t really summarized in the same way as, say Bee or Plaud. Some may appreciate the brevity, but I found relying on Limitless to be a lot of work. Of course, you can use the chat tab to ask the Limitless AI to find and summarize anything it has recorded.
Unlike the other apps, which present each day one after the other in an infinite scroll, Limitless puts each day on its own page. You can swipe pages to change days, or tap a calendar to zip right to a specific date.
The two discs stick together magnetically, so you can clip the pendant over clothes—but if you prefer, you can slip it onto a lanyard and wear it around your neck.
Dave Johnson For Forbes
The app has a ton of customization features—language, a custom word dictionary, how long to keep audio clips and much more. In theory, it also learns your voice, so it knows when you are speaking and can pick you out of a group conversation. You can also teach it other voices it’ll commonly hear, like coworkers, friends and family. All that said, I didn’t see it clearly understanding it was me when I was speaking throughout testing—but hopefully it’ll get better at that over time.
Limitless gets pricey, though. The pendant has a one-time cost of $199 and comes with 20 hours per month of usage for free. You’ll probably burn through that pretty fast, though. It’s designed to be on all day long. A better bet is the Pro subscription (100 hours per month) for $19 per month, or the unlimited plan for $29 per month.
Dimensions: Button shaped, 1.1 x 0.4 inches | Charging: Proprietary dock | Data and Security: Not HIPAA compliant, encryption at rest and in transit, secure storage | Wear style: Lanyard | Recording mode: All day/lifelogging | Battery life: One day
Pros:
Exciting ecosystem of apps to process your recordings in different ways
Learns voices to distinguish speakers
Automatically collects relevant to-dos from your conversations
Always on (unless you mute it)
Cons:
Omi transitioning from free service to $20/month subscription
Reasonable concerns about the privacy and security of the apps and integrations
I find Omi utterly fascinating. While the app is similar to Bee—it summarizes your conversations while also extracting to-dos and various facts—it is the most future-facing of the AI pendants you can buy today. That’s because Omi’s parent company, Based Hardware, has built an open ecosystem of apps—essentially, plugins for the Omi app—that let you synthesize and manipulate your recordings in all sorts of cool and unexpected ways.
In addition to the standard summarizer, there are hundreds of other options you can select, such as a cognitive bias detector (which gives recommendations for more objective thinking), a mentor app that dispenses advice about your conversations, a “lie detector,” a joke extractor that sniffs out humor in your day and many others. Don’t expect all of these apps to be gems, but the fact that Based Hardware is laying the groundwork for this kind of integration is really exciting. On the other hand, to use these apps, you need to give third party developers access to your personal data, which could be alarming to some.
The Omi pendant is shaped like a fat coin (about a half-inch thick) and is meant to be worn around your neck on a lanyard. A button starts or stops recording, but you can leave it running all day long.
Omi’s summaries are good, but the availability of additional plug-in apps gives you a lot of flexibility.
Dave Johnson For Forbes
The UI is a bit more minimalist than some of the competing AI pendants, but even so, the summaries are fun to read. Each one starts with a broad summary (and weirdly, the greeting, “hey now”), along with sections for the mood of the activity, action items, key takeaways, questions raised and funny or ‘otherwise’ notable quotes (like when Omi caught my fiancé in a rare moment admitting an error, saying: “‘You’re right. We should’ve turned right.’ — Beth.”)
In keeping with the more minimalist approach, transcripts are a little harder to get to. There’s no way to see them directly in the app, though you can use a menu to export them, such as via email. Need more information than you can find in the summaries? Use the AI chat.
The Omi pendant is shaped like a fat coin (about a half-inch thick) and is meant to be worn around your neck on a lanyard.
Dave Johnson for Forbes
I really like the fact that Omi confidently creates a pretty effective to-do list without forcing me to approve each one, like the overly-cautiously Bee. On the other hand, Omi’s fact collector—which seems similar to Bee’s—is weird in a different way. Its facts are referred to as “memories,” and there’s no apparent rhyme or reason for why Omi selects the ones that it does. Memories are a crazy combination of things I’ve personally said, other people have said, and things it overheard media saying. Aarav Garg, one of Omi’s founders, told me that the long term plan is for Omi to use these facts to build sophisticated understanding of the user and its environment, but they don’t appear to do much, if anything, at the moment. Moreover, like Limitless, Omi promises to learn both your own voice and those of others, though in my experience Omi couldn’t reliably tell me apart from others in practice.
AI Recording Wearables, In Summary
There are already no fewer than four AI wearables on the market, giving you a lot of choice if you want to clip a recorder to your shirt and let artificial intelligence keep track of your day for you. And while the tech is still in its infancy—ChatGPT debuted to the public less than three years ago, and these are all first-gen AI pendants—they’re quite good. I thoroughly enjoyed testing them, and will continue to use them (especially Plaud and Omi) long after this article has been published.
If you’re looking to dabble in AI recording and mostly want to use a wearable to keep tabs on your personal life, you can’t go wrong with Bee. But if you want to be better organized at work, I suggest that you try one of the other three. I’m most excited about Plaud, but it has the disadvantage of requiring you to start and stop recording for each meeting; Omi and Limitless work in the background all day long.
None of them work perfectly, and the most common problem I encountered was they all either confused who was speaking or didn’t even try to tell everyone apart. Many of my summaries confused me with my co-workers, or simply said “someone suggested” a particular idea. Is that a problem? Six months from now, I’ll have forgotten the context of the conversation and this kind of ambiguity could at times be confusing. But a few days after the conversation, this is a minor inconvenience, and the summaries are super helpful.
It’s not unreasonable to have concerns about data privacy, security, and the pervasiveness of everyone being recorded all the time, often without their knowledge. This gives me a little pause as well. Even so, AI recording feels like the future to me, and I would not be surprised if lifelogging—i.e., all-day AI recording—were common, routine and mundane as soon as five years from now. In fact, these devices might be the iPods of the 2020s; eventually, AI recording with be built into our smartwatches, smartphones and smart glasses, making standalone AI pendants as quaintly obsolete in 2030 as the iPod Nano is today.
Why Trust Forbes Vetted
The Forbes Vetted team has published hundreds of tech and electronic guides that leverage our hands-on testing and research, including audio stories, like the best outdoor speakers and more.
This story was written by executive editor Dave Johnson, a veteran tech journalist who has authored nearly three dozen books on consumer technology and digital photography.
He’s obsessed with tools and tech related to generative AI, and is the author of our roundup of the best AI writing tools. For this story, Johnson tested all the wearable devices for at least five weeks.
When he’s not testing AI wearables, Johnson shares the latest insights on the best pizza ovens after making over 150 pizzas. He has also ridden hundreds of miles to find the best electric bikes.
This article weas edited by consumer tech and electronics editor Rebecca Isaacs, who has been involved in tech journalism for over six years. During her time at Forbes Vetted, she’s tested many wearables, from the Ray-Ban Meta Smart Glasses to the Apollo Wearable and more. When she’s not writing about the best smartwatches for women, she’s testing most types of audio equipment, including wireless headphones, wireless earbuds and Bluetooth speakers.
Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
The Androidify app lets you customize your own Android bot.
You can use a photo or enter a text prompt.
There’s an app and a browser version of the tool.
If you’re an Android die-hard, Google is giving you a fun new way to express your fandom.
For more than 15 years, a small green robot (a.k.a. Bugdroid) has been the official mascot of the Android operating system. Several versions of the mascot have existed, with the most recent 3D version rolling out in 2023.
Google debuted a “Create Your Own Android Bot” tool last year that let you dress up your own bot with clothing and accessories, but it was limited. The company has a new tool now, though, and it’s anything but.
Also: I tried every new AI feature on the Google Pixel 10 series – my thoughts as an AI expert
The new Androidify app lets you make your own bot avatar, but it does so based on your own photo or text prompt, meaning it has an incredibly deep level of customization.
Google says the app uses a combination of tools to create your bot — Gemini 2.5 Flash to caption the photo, Imagen to generate your custom Android bot, and Veo 3 to animate your bot with different vibes.
I made my own Androidify bot
I tried several photos and quickly found how fun it was to see what the output looked like. I had the most success with a Halloween selfie from last year where I was wearing a Ghostbusters jumpsuit and proton pack. My Android bot (green, of course) had his own tan jumpsuit, proton pack, and even a version of the “no-ghost” logo on his sleeve.
The details weren’t perfect, but the overall vibe was very fun. I tried several more unique photos, and each one came out just as good.
Also: Your Gemini app just got a major AI image editing upgrade – for free
I wasn’t quite able to get the same level of detail from the text prompt no matter how specific I was, but the prompt bots still captured the feeling pretty well.
Screenshot by Artie Beaty/ZDNET
How to make your own Androidify bot
To make your own bot, download the Androidify app (you can also use the browser version).
You can choose between using a photo as the basis for your bot or a text prompt to describe what you want.
Also: Did Google just give us the ultimate AI photo-editing tool? I tested it on the Pixel, and hard agree
If you go the photo route, you can either take a photo or pick one from your gallery. Choosing from your gallery is a bit more fun because you can choose a photo with accessories or a unique outfit, which the app will replicate.
It’s better to choose a photo with one person because you don’t get to pick which one you want. If there are multiple people, the app will either pick one person or create a strange mashup of the two people, like it did with a photo of me and my daughter (it took my hair and facial hair and slapped it on a bot wearing her dress).
Also: I used Google’s photo-to-video AI tool on my selfie – and it made me do the tango
Once you pick your photo and tap “Transform,” the app gets to work creating your bot. When your bot is done, you can choose from several formats — a large phone wallpaper, a banner, a sticker, or a 3:1 size that would serve as a social media cover photo. From there, you’ll choose your background from one of 15 choices.
VIDAA has partnered with waipu.tv to make the TV streaming service available on all smart TVs in Germany using its operating system.
The move means viewers can access waipu.tv without additional hardware.
The waipu.tv service offers more than 300 live TV channels, including over 70 pay-TV channels, alongside cloud-based recording, pause and restart functionality, and a library of 40,000 on-demand titles within its video-on-demand section, the waiputhek.
“Germany is one of the most competitive and innovative TV markets in Europe, and waipu.tv is a perfect example of that,” said Guy Edri, CEO of VIDAA. “This partnership brings together great technology, great content, and great user experience – all on a single smart TV platform. We’re proud to make it available, out of the box, to millions of homes across Germany.”
Markus Härtenstein, Co-CEO of Exaring, the operator of waipu.tv, added: “At waipu.tv, our vision is to radically simplify television. Partnering with VIDAA is another step in bringing this vision to life for millions more viewers on the big screen – seamlessly integrated and just one click away.”
The integration strengthens VIDAA’s German content line-up and reflects its strategy of building deeper partnerships with regional and national broadcasters, according to a statement, adding that viewers will benefit from personalised TV recommendations, quick access to favourite channels and recordings, and smooth playback through its performance engine.
The waipu.tv app is available from today and comes preinstalled on VIDAA-powered televisions sold in Germany, including models from Hisense, Toshiba TV, Dyon, Loewe and Leica.
