After a couple years of having its open-source Llama AI model be just a part of its Connect conferences, Meta is breaking things out and hosting an entirely generative AI-focused developer conference called LlamaCon on April 29. The event is entirely virtual, and you'll be able to watch along live on the Meta for Developers Facebook page.

LlamaCon kicks off at 1PM ET / 10AM PT with a keynote address from Meta's Chief Product Officer Chris Cox, Vice President of AI Manohar Paluri and research scientist Angela Fan. The keynote is supposed to cover developments in the company's open-source AI community, "the latest on the Llama collection of models and tools" and offer a glimpse at yet-to-be released AI features.

The keynote address will be followed by a conversation at 1:45PM ET / 10:45PM ET between Meta CEO Mark Zuckerberg and Databricks CEO Ali Ghodsi on "building AI-powered applications," followed by a chat at 7PM ET / 4PM PT about "the latest trends in AI" between Zuckerberg and Microsoft CEO Satya Nadella. It doesn't seem like either conversation will be used to break news, but Microsoft and Meta have collaborated before, so anything is possible.

Meta hasn't traditionally waited for a conference to launch updates to Meta AI or the Llama model. The company introduced its new Llama 4 family of models, which excel at image understanding and document parsing, on a Saturday in early April. It's not clear what new models or products the company could have saved for LlamaCon.

LlamaCon will stream live on April 29th through the Meta for Developers Facebook page and we'll be live-blogging the event right here on Engadget so you can get all the details as they happen.

Android tablet or foldable owners can now adjust the divider between Gmail's message list and conversation views. 9to5Google spotted the change, which arrived in version 2025.04.13.x of Gmail for Android.

The update lets you drag your finger on the divider to dynamically change the sizes of the Gmail app's two panes: the list on the left with multiple messages and the pane on the right that shows individual emails and threads. Alternatively, you can drag the boundary to the display's edge to show only one or the other.

The adjustable divider is also available in the Google Chat app for large-screen Android devices. The new feature is available to everyone with a personal Google or Google Workspace account in both apps.

Almost a year since Microsoft announced its controversial Recall feature, and after several delays, the company has finally started bringing it to Copilot+ AI PCs today. The launch comes just a few weeks after Microsoft started testing Recall broadly with Windows Insiders. There are also a few other AI-powered features coming along with this release, including an improved Windows Search and Click to Do, which lets you quickly use AI features from within your existing apps. As usual, the release won't immediately roll out to all Copilot+ PCs, instead Microsoft is gradually releasing it over the next month (and likely monitoring potential issues along the way).

Recall was one of the biggest announcements at Microsoft's Copilot+ debut last May, but almost immediately, it came under fire for some glaring privacy issues. At a basic level, Recall constantly records what you're doing on your PC via screenshots, and it uses AI to search them for specific words and images. The idea is that you'll never forget where you put a document you were working on weeks ago, or which random website you've lost track of. Security and privacy advocates were initially concerned that Recall was automatically enabled on Copilot+ PCs and that it wasn't storing its database of screenshots securely. That led to an immediate delay for Recall that lasted for several months. 

In November, Microsoft finally revealed how it will make the feature even more secure. Its snapshots and related data will be stored in VBS enclaves, which the company describes as "software-based trusted execution environment (TEE) inside a host application." Additionally, you'll have to turn Recall on manually when you set up a Copilot+ machine, it will rely on Windows Hello biometric security to make any settings changes, and it can be completely uninstalled if you want to be rid of it entirely.

While it's heartening to see Microsoft take security more seriously after all of Recall's initial criticism, it's still worrying that it took widespread condemnation for any of it to happen. The company's rush to deliver a shiny new AI feature to sell Copilot+ PCs, and snub the likes of Google and Apple, ultimately got in the way of delivering the best product for consumers. It'll be hard to trust Recall, or really any of Microsoft's AI-enabled Copilot features, because of its initial blunder.

Less controversial is the improved Windows Search, which will let you find documents and images in your own words. That means you shouldn't have to worry about remembering specific file names or other minutia to find what you need. Like all of the Copilot+ features, including Recall, the improved search runs locally using the neural processing units (NPU) in AI PCs. There's nothing being sent to the cloud.

I'm personally the least excited about Click to Do, but there may be an audience for people who want easy access to Microsoft's AI tools. You'll be able to highlight text and quickly have it summarized or rewritten by Copilot, without dumping it into the Copilot app specifically. You can enable the feature by pressing the Windows key and clicking on your screen, swiping right on a touchscreen or hitting the Click to Do icon as it pops up throughout Windows (you'll see it in places like the Start menu and Snipping Tool).

Microsoft says Click to Do actions for images are available on all Copilot+ PCs with the new Windows 11 April update, and text actions will be available on Snapdragon systems today, and eventually on Intel and AMD AI PCs.

I've briefly used all of these features on a Surface Pro Copilot+ machine using the latest Windows 11 Insider build, but I've been waiting to test their official release before making any final judgements. I can say that Recall mostly works as advertised — it was easily able to bring up a document I was viewing a week later, and it quickly found a few websites I was viewing — but it also didn't add much to my Windows experience. At this point I religiously save websites I need to revisit via Pocket, and I'm well-versed enough in Windows to know where I've put my files. Recall isn't really made for me, though, it's for less experienced users who just want to find their stuff.

Even power users will like the improved Windows search, though, but that's only because the platform's search has always been notoriously awful. And while I'm not a huge proponent of AI text summarization, but Click to Do did a decent job of summarizing a few long articles.

AI company Perplexity announced in February that it was building its own browser called Comet. In a recent interview with the TBPN podcast, CEO Aravind Srinivas gave some insight as to why the business appeared to be branching out from its artificial intelligence focus: It's to collect user data and sell them targeted advertisements.

"That’s kind of one of the other reasons we wanted to build a browser, is we want to get data even outside the app to better understand you," he said. “We plan to use all the context to build a better user profile and, maybe you know, through our discover feed we could show some ads there.”

If that all sounds familiar, it could be become Google's Chrome browser has taken a similar approach. In fact, Comet is built on Chromium, the open-source browser base from Google. That's not to say Perplexity wouldn't take the chance to go straight to the source and acquire Chrome in the aftermath of Google's recent monopoly court ruling regarding online search. In the ongoing hearings about Google and its potential sale of Chrome, Chief Business Officer Dmitry Shevelenko said he thought Perplexity would be able to continue running the browser at its current scale. Unsurprisingly, he wasn't too keen on OpenAI acquiring the property.

Back when Meta first introduced its Twitter competitor Threads, many noted that the company had failed to secure the threads.com domain and instead launched the website at threads.net. At the time, the Threads dot com domain belonged to a messaging app startup that said it was reluctant to rebrand its business.

But that startup was later acquired by Shopify and Meta did eventually acquire the coveted threads.com domain for an undisclosed amount. Now, Meta is finally moving Threads’ website to threads.com, and adding some much needed functionality to the web version of Threads.

The update adds a new composer that pops up in its own window so you can continue to browse your feeds as you type out a new post. It also allows you to scroll your various custom feeds in a single-column view (much like Threads’ mobile app), and finally adds a menu shortcut for saved posts. (Previously, the only way to view saved posts on web was to add it as a pinned column.)

Screenshot via Threads

Meta is also stepping up its efforts to lure users directly from X. The company says it’s testing a new feature that allows users to upload a list of people they follow on X and find the corresponding accounts on Threads. The feature, currently labeled as being in “beta,” sounds a bit clunky according to Meta’s in-app description. It notes that downloading data from X can take as long as three days, so it’s not exactly a simple process. But in addition to giving users a way to find familiar accounts on Threads, it could also give Meta some valuable insight into users’ habits on other platforms.

The Spacetop AR laptop made a splash when it debuted a few years ago with an intriguing pitch: What if you could have a notebook that works entirely through augmented reality glasses, without a built-in screen of its own? Unfortunately, we found the Spacetop experience to be underwhelming, and the hardware seemed like a tough sell for $1,900. Last Fall, Spacetop's creator Sightful told CNET that it was abandoning the screen-less laptop altogether and instead focusing on building AR software for Windows PCs. Now, we have a clearer sense of what Sightful is up to. 

Today, Sightful is officially launching Spacetop for Intel-powered Windows AI PCs, following a short trial launch from January. For $899 you get a pair of XREAL's Air Ultra 2 glasses and a year of Spacetop's software. Afterwards, you'll have to pay $200 annually for a subscription. The software works just like the original Spacetop concept — it gives you a large 100-inch AR interface for doing all of your productivity work — except now you're not stuck with the company's middling keyboard and other hardware.

Spacetop might be compelling to AR fanatics who already own Intel AI PCs like the Surface Laptop Pro for Business, or a Dell XPS machine with the Core Ultra 7 chip. But if you don't have one of those computers, you'll have to add it to Spacetop's already high cost. And don't forget about prescription lenses: Those will run you $50 for single-vision and $150 for progressive.

Spacetop doesn't support Intel chips without NPUs, as its AR interface requires constant AI processing. It doesn't work AMD or Qualcomm's AI CPUs, either. The company partnered with Intel to optimize Spacetop for its platform, which included help with understanding Intel's NPU architecture as well as tweaking its chips to help the AR software run better. In a conversation with Engadget, Sightful CEO Tamir Berliner noted that the company might pay more attention to other chip platforms if it gets similar attention. 

As for Mac support, Berliner said in a statement to Engadget: "Yes, we plan to create a version of Spacetop for Mac — the timing is still TBD based on a lot of different factors regarding laptops, silicon, compute power, glasses, and more. We will be excited to share more news when we are able.”

Given its price and the strict PC requirements for using it, Spacetop clearly isn't meant for a wide audience. But there are potential benefits for early adopters. Its large AR screen could make it easier to multitask without large monitors, and it's also easier to focus on your apps since Spacetop vastly simplifies the Windows 11 interface. There are some obvious privacy benefits, too, since no one else can see what you're looking at in AR.

The big downside, though, is that you'll have to get used to wearing Xreal's large Air 2 Ultra glasses. When we demoed it at CES, we found it to be an improvement over previous Xreal frames, thanks to their sharp 1080p micro-OLED displays and wider field of view. The Air 2 Ultra are also notable for having 6DoF tracking, which allows you to move around AR objects. While sleeker than the Vision Pro, the glasses are still pretty clunky, and you'll also have to snap in additional prescription frames if necessary.

I'll need to see this latest iteration of Spacetop in action before making any final judgments, but it's clearly a more viable concept as an app that can work on a variety of laptops. Nobody wants to buy bespoke hardware like the old Spacetop laptop, no matter how good of a party trick it may be.

Perplexity has rolled out an update for its iOS app, giving iPhone users access to its AI voice assistant that was initially released for Android users earlier this year. Its voice assistant can perform tasks for the user by browsing the web or accessing other apps for them. If they ask the assistant to find them a table for a specific restaurant, for instance, Perplexity can launch the OpenTable app with the number of people, the date and the time already filled out. The user still has to perform the final action and book a reservation, but it's already laid out for them — all they have to do is click the button. 

Users can also ask the assistant to draft emails for them for specific contacts, which they'll have to send themselves, and create reminders for them on the calendar. They can ask it to recommend them spots they could visit, such as restaurants serving food they want to eat, and Perplexity will mark locations on the map. Of course, they can ask the assistant to do web searches for them, such as finding specific videos that Perplexity can open on the YouTube app. 

As The Verge notes, Perplexity's voice assistant works on older iPhones, unlike Apple Intelligence that only works on the company's more recent models. The iOS assistant also doesn't have the capability to look at the world for the user and tell them what they're seeing yet, though the feature is already available on Android and could make its way to iPhones in the future. 

f

Strava has had a few new developments recently, including buying personalized running plan app Runna. Now, it has launched a new training feature for distance runners called Performance Predictions, which gives estimated finish times for difference race lengths. 

The app will now offer possible times for a 5K, 10K, half marathon and marathon — so, now, it's not only your own goals you have to live up to but Strava's as well. No pressure. Strava is providing these predictions using a machine learning model (shocking) which looks at over 100 data points from the individual and the performance of similar runners on the app. The times should change after every run and based on rest periods. 

Strava claims that users logged almost 1 billion runs on its app last year, so it should be helpful to some (and apply added pressure to others). Subscribers can access it in Strava's Progress tab. It will also show you how much those numbers have recently changed based and your average time per mile. 

If you're a frequent Photoshop user, there's a good chance you've run into this scenario. You open the program after a long break to edit an image, but this being Photoshop we're talking about, there are about five different ways to complete the task before you and you can't quite remember the way you learned to do it. Adobe is trying to make it easier to use its flagship app with the introduction of a built-in AI agent that can navigate Photoshop and complete tasks for users. At its Adobe Max London event today, the company demoed this agent, showing how it can automate multi-step workflows.      

Users can access the tool from the redesigned Actions panel. If you've used an AI chat bot before, the interface will be familiar. There's a text box for users to input what they want the agent to do for them, with a list of suggested prompts above. Once you hit enter, the tool will display all the steps needed to complete your request, allowing you to follow along as it does the work for you. Judging by the demos Adobe shared with press before the event, the agent can complete nearly any task you might turn to Photoshop to do. In one example, the tool first applies a color gradient to text, and then does the same for the background behind it. 

Adobe

At the same event, Adobe showed off a new version of its Firefly app, which brings together all of the company's AI image, video, audio and vector generation tools in one easy to find place. The redesigned Firefly is available to use on the web today, with Android and iOS apps coming soon. Additionally, each part of the app is powered by new underlying models that offer better performance. 

With image generation, for instance, Adobe is offering two new in-house systems, the imaginatively named Firefly Image Model 4 and Firefly Image Model 4 Ultra. Of the former, the company says it can produce 2K resolution images, making it possible to print what the model generates. Adobe claims both systems offer best-in-class human rendering. All of Adobe's own models are commercially safe, meaning they weren't trained on copyrighted material, and creative professionals can feel safe using them. 

"By unifying image, video, audio and vector generation and providing unmatched creative control, Firefly empowers creative professionals to work more productively and with an unmatched degree of precision," Adobe said. "Seamlessly integrated with Photoshop, Premiere Pro, Express and Adobe’s other creative applications, it offers AI-powered assistance throughout every stage of the content creation process – from ideation through production."

Adobe

If Adobe's models aren't thing, the company is also — for the first time — offering third-party models directly within the Firefly app. With today's announcement, some of the more notable options include Google's Imagen 3 and Veo 2 models, as well as ChatGPT image generation, with more to come later. As part of Adobe's agreements with Google, OpenAI and other model providers, those companies have agreed to not use data from Adobe users for training their future AI systems.   

WhatsApp is introducing a new feature called Advanced Chat Privacy that can block participants from sharing the contents of a conversation. This is an opt-in tool that’s available for both traditional chats and groups.

Once selected, the toolset will prevent anyone in the chat from sharing anything outside of the app. This means it’ll block all chat exports, but that’s just the beginning. The software also prevents a smartphone from auto-downloading media and will stop integration with AI assistants. Basically, what happens on WhatsApp stays on WhatsApp. However, it's unclear if it prevents screenshotting. We've reached out to Meta and will update this post when we hear back. 

The platform says this is "best used when talking with groups where you may not know everyone closely but are nevertheless sensitive in nature." It gives examples like a support group about health challenges and a chat about community organizing.

WhatsApp says this is just the first version of the tool and that it’ll be adding "more protections" in the future. It’s rolling out right now across the globe, though it could take a month or two to reach everyone. 

The platform is already known for being privacy-focused, as it uses end-to-end encryption. It also leverages related software tools like disappearing messages and chats that require device authentication.