🚀 The M&A Race Heats Up: Enterprise IT is in the midst of a consolidation wave. Companies are reassessing their investments and mergers & acquisitions (M&A) activity is booming. Why? The age of AI has arrived, and everyone wants a piece of the pie. But here’s the catch – so far, AI’s impact has been underwhelming. A recent MIT study found that only 5% of enterprise AI rollouts have delivered meaningful value. Ouch!

🤔 Are We Heading in the Right Direction? Today’s M&A strategies often focus on acquiring a complete AI stack for IT teams. But is that the right approach? Not quite. True value won’t come from just reorganizing legacy apps for tech-savvy users. It’ll come from empowering business users and analysts – the people closest to the work – to reimagine their processes in an AI-driven world.

🌊 The Lakehouse Dilemma: Many enterprises have migrated data into modern lakehouse architectures, enabling centralized analytics. But AI brings new challenges. Connecting AI models directly to vast stores of sensitive data is a governance nightmare. The solution? Give AI access only to the relevant data needed for each specific use case. But that’s not all – data in lakehouses is often shaped by the enterprise applications it came from. It needs to be made usable by AI, and that means embedding the business logic that underpins day-to-day processes.

🌐 The Rise of the AI Data Clearinghouse: Enter the AI Data Clearinghouse – a neutral, business-friendly software layer that connects disparate systems and allows business users to design AI workflows visually, with built-in governance and process logic. This resonates with business leaders because it addresses the friction points stalling enterprise AI. It democratizes AI process creation, speeds up deployment, and makes it easy for executives to understand and approve use cases.

🔒 Unlocking AI’s Potential: For CEOs hesitant to feed first-party data into AI, the clearinghouse offers a middle ground. It turns AI from a mystery box into a transparent enabler of decision-making and collaboration. Without this approach, AI will remain stuck in pilots and proofs of concept, never scaling to real impact. So, let’s not let MIT’s findings gather dust. Let’s act now and draw meaningful value from our AI investments.

💡 Empowering Business Users with Data: Too many vendors are pitching data platforms and copilots as the fast track for IT teams to bring AI success to the business. But IT can’t do it alone. Embedding AI across organizations means putting intuitive, governed AI workflow tools directly in the hands of business users. When those tools serve the dual purpose of embedding compliance guardrails, leadership can have confidence that AI is being deployed responsibly.

🏆 The Future of AI in Enterprise: As AI moves from experimentation to enterprise-wide adoption, the winners will be the organizations willing to rethink both their data architectures and their assumptions about who owns AI. By embracing the clearinghouse model, businesses can unlock the next wave of value: AI that is transparent, trusted, and driven by the very teams closest to the customer and the work.

🚨 Attention, Windows users! Microsoft has just dropped a cryptic teaser that’s got us all in a flutter. They’re promising “something big” for Windows 11, and it seems like our trusty keyboards might get a break! 👋

The official Windows account on X posted a mysterious message about our fingers getting some well-deserved R&R, and that we should get ready for a major announcement on Thursday, October 16. 📅

Now, we’ve all been speculating, and it looks like voice commands might be the game-changer here. Microsoft’s own execs have been hinting at this for a while now. David Weston, their Corporate VP for OS Security, even predicted that the future of Windows will be all about talking to our PCs, not just typing and clicking. 🗣️💻

And Pavan Davuluri, the VP of Windows and Devices, told us that we’ll soon be able to chat with our computers while we’re multitasking or collaborating with others. 🤝💬

So, could this be the big reveal? Will Windows 11 take a giant leap towards hands-free computing? We’ll have to wait and see, but one thing’s for sure – Microsoft’s been making some serious strides with voice commands already. From Voice Access to voice dictation in Word, they’re not messing around.

Could this new feature be tied to AI and require some serious on-device processing power? Only time will tell, but we’re on the edge of our seats waiting for Thursday! 🎬

Stay tuned, folks! And if you’re not already, make sure to follow TechRadar for all the latest tech news and updates.

Get ready, racing fans! The National Hunt season is about to kick off, and we’ve got the ultimate guide on how to watch it all for free, no matter where you are in the world. So, grab your binoculars and let’s dive in!

Season Opener & Key Dates
– The season begins in Carlisle on October 16, with the action unfolding over seven months until late April/early May.
– Mark your calendars for the major festivals: Cheltenham (Nov, Jan, Mar), Aintree (Apr), and Punchestown (May).

Stars Returning for 2025/26
– Expect to see top horses like Jonbon, Constitution Hill, Galopin des Champs, and Inothewayurthinkin gracing the tracks this season.

Watching on ITVX
– ITVX, the UK’s online streaming service, will broadcast select jump races, including Cheltenham, the Grand National, and live Saturday racing.
– No subscription needed! Just a valid TV license and an email account.
– First-time users can register for free or download the app.

Watching Abroad
– If you’re traveling or working outside the UK, you’ll need a VPN to access ITVX’s coverage. NordVPN is our top pick for reliability and ease of use.
1. Install NordVPN.
2. Connect to a UK server.
3. Enjoy the races on ITVX!

ITVX’s Coverage
– Expect live races, highlights, and a full program, with expert analysis from Ed Chamberlin, Francesca Cumani, Mick Fitzgerald, Alice Plunkett, and Luke Harvey.
– ITVX is available on various devices, including Amazon Fire, Android, Apple TV, Google TV, and more.

Carlisle Season Opener
– The season begins with a bang at Carlisle, where new stars emerge and strategies unfold.
– Keep an eye on trainers Olly Murphy, Donald McCain, and Paul Nicholls, and jockey Harry Atkins.

Full National Hunt Meeting Schedule
– Click to see more ▼

Don’t Miss Out!
– Stay tuned for more updates and enjoy the thrilling 2025/26 National Hunt season!

Get ready to level up your workout game, tech enthusiasts! The newly announced Oakley Meta Vanguard glasses have taken the sports tech world by storm, promising real-time stats and hands-free interaction with your Garmin device. But which Garmin gadgets can truly unlock the Meta Vanguard’s potential? We’ve got the scoop!

The Meta Vanguard’s Game-Changing Features
These sleek specs sync with your Garmin device to provide real-time stats like heart rate and pace, all without lifting a finger. Plus, they can automatically capture milestones from your run to share on your favorite social media platforms, including Strava.

But Wait, Not All Garmin Devices Are Compatible…
In our quest to test the Meta Vanguard, we initially grabbed a Garmin Instinct 3, only to find it wasn’t as compatible as we’d hoped. Don’t make the same mistake! We’ve now got the full list of compatible Garmin devices, so you can make an informed decision.

Compatible Garmin Devices: Smartwatches & Bike Computers
Here’s the lineup, ranging from budget-friendly to premium options:

Smartwatches:
– D2 Mach 1, D2 Mach 1 Pro
– Enduro 2, Enduro 3
– Epix Gen 2, Epix Pro Gen 2
– Fenix 7, Fenix 7S Pro, Fenix 7X Pro, Fenix 8, Fenix 8 Pro, Fenix E
– Forerunner 165, Forerunner 265, Forerunner 265S, Forerunner 570, Forerunner 965, Forerunner 970
– Marq (Gen 2) Adventurer, Marq (Gen 2) Athlete, Marq (Gen 2) Aviator, Marq (Gen 2) Captain, Marq (Gen 2) Carbon Edition, Marq (Gen 2) Commander, Marq (Gen 2) Golfer
– Quatix 7, Quatix 7 Pro, Quatix 7XT
– Tactic 7, Tactix 8, Tactix 8 Solar
– Venu 3, Venu 3S, Venu 4, Venu X1
– Vivoactive 6

Bike Computers:
– Edge 540, Edge 550, Edge 840, Edge 850, Edge 1040, Edge 1050, Edge MTB

Which Garmin Should You Buy?
If you’re in the market for a new Garmin watch to pair with your Meta Vanguard glasses, consider these top picks:

1. Garmin Venu X1 – Our top-rated fitness tracker, perfect for both serious athletes and casual enthusiasts.
2. Garmin Fenix 8 – A premium option with advanced GPS, a new heart rate sensor, and impressive battery life.
3. Garmin Fenix 8 Pro – An upgraded version of the Fenix 8 with LTE connectivity and satellite communication.
4. Garmin Forerunner 165 – Our favorite budget Garmin watch, offering great value for money.

Don’t miss out on the ultimate sports tech combo – grab your Oakley Meta Vanguard glasses and pair them with one of these compatible Garmin devices today! 🏃‍♂️🚴‍♀️🏂

Alright, gamers, buckle up! Microsoft’s Xbox Game Pass Ultimate prices are about to take a leap next month, and many of you are feeling the pinch. The announcement has already sent waves of cancellations, with the official site struggling to keep up. But don’t throw in the towel just yet! Amazon’s got a sneaky little deal that might just save your gaming budget.

Starting November, the Ultimate tier’s monthly fee is jumping to $29.99. That’s a hefty increase, and it’s left many subscribers feeling sick to their stomachs. If you’re one of them and you’re ready to call it quits, we’ve got a guide on how to cancel your subscription. But if you’re not quite ready to part ways with Xbox Game Pass Ultimate, Amazon’s got your back.

Right now, Amazon’s slashed the price of a 3-month Xbox Game Pass Ultimate subscription to just $54.99 (down from $59.99). It’s not a massive discount on the surface, but when you consider that the same amount of subscription time will cost around $90 after the price increase, you’re looking at a sweet $35 saving!

Today’s HOT DEAL: With Xbox Game Pass prices skyrocketing next month, this could be your last chance to save big on three months of subscription time. It’s already the #1 best seller in Amazon’s ‘Xbox Games, Consoles & Accessories’ category.

1-month subscription: $19.99 at Amazon – View Deal

Microsoft’s not new to price hikes, and we’ve seen some significant increases over the past year. Once the new prices kick in, you’ll be spending around $360 a year on Xbox Game Pass Ultimate. Pair that with the cost of an Xbox Series S console ($379.99), and you’re looking at a total of around $739 – just $10 cheaper than a PS5 Pro.

After the price increase, I can’t in good conscience recommend subscribing to Xbox Game Pass Ultimate. But for now, if you’re in the US and you want to keep your subscription going, this Amazon deal is your best bet. Especially if you’re waiting for Call of Duty: Black Ops 7 to drop on the service on November 14.

So, what are you waiting for? Grab this deal while you can, and keep your gaming marathon going strong! And don’t forget to follow TechRadar on Google News, TikTok, and WhatsApp for more gaming news and updates. Happy gaming!

Oracle has swiftly patched another critical vulnerability, CVE-2025-61884, in its E-Business Suite. This fix comes hot on the heels of reports that the notorious hacker group, ShinyHunters, allegedly exploited this flaw to pilfer sensitive corporate data from multiple organizations. This is Oracle’s second patch in recent weeks addressing exploit chains used in ransomware extortion campaigns.

The newly patched vulnerability, discovered in Oracle’s E-Business Suite, is a doozy. It’s remotely exploitable without any authentication, meaning hackers can target it over the network without needing a username or password. If successfully exploited, it could grant access to sensitive resources within the system.

The affected versions range from 12.2.3 to 12.2.14. Oracle, as always, urges customers to stay on actively-supported versions and apply all Security Alerts and Critical Patch Update security patches promptly.

While Oracle’s advisory doesn’t mention ShinyHunters or the recent breaches, our tech-savvy friends at BleepingComputer confirmed that this patch indeed breaks the exploit chain used by the threat actors. This is the second time in recent weeks that Oracle has patched E-Business Suite vulnerabilities allegedly exploited by cybercriminals.

Earlier in October, businesses across the U.S. received extortion emails from ransomware actors known as Cl0p. Initially, Oracle claimed the attackers were exploiting an n-day vulnerability patched earlier. However, it later released a patch for CVE-2025-61882, a bug that allowed unauthenticated attackers to compromise and fully take over the Oracle Concurrent Processing component of E-Business Suite.

Now, with the second patch out, let’s hope these security holes are finally plugged. Stay tuned for more updates!

Ever wondered how your AI agent’s performance would skyrocket if it could learn purely from its own actions, without any rewards or human demonstrations? Meta Superintelligence Labs has just made that a reality with ‘Early Experience’, a revolutionary reward-free training approach that’s leaving imitation learning in the dust across eight diverse benchmarks.

What’s Early Experience all about?

Traditional AI pipelines rely heavily on imitation learning (IL) from expert trajectories, which is cheap but hard to scale and unreliable when faced with new situations. Reinforcement learning (RL), on the other hand, promises learning from experience but requires verifiable rewards and stable infrastructure, which aren’t always available, especially in web and multi-tool settings.

Early Experience sits right in the middle, offering the best of both worlds. It’s reward-free like IL, but the supervision comes from the consequences of the agent’s own actions, not just expert actions. In simple terms, the agent proposes, acts, and learns from what actually happens next—no reward function required!

The team behind Early Experience has come up with two concrete strategies to make this happen:

1. Implicit World Modeling (IWM): The model learns to predict the next observation given the state and chosen action, helping the agent tighten its internal model of environment dynamics and reduce off-policy drift.

2. Self-Reflection (SR): The model explains why an expert action is better than alternatives using observed outcomes, then fine-tunes the policy from this contrastive signal.

Both strategies use the same budgets and decoding settings as IL, but with a twist—the data source is agent-generated branches, not more expert trajectories.

Putting Early Experience to the test

The research team evaluated Early Experience on eight language-agent environments covering web navigation, long-horizon planning, scientific/embodied tasks, and multi-domain API workflows. The results? Early Experience delivered average absolute gains of +9.6 success and +9.4 out-of-domain (OOD) over IL across all tasks and models. These gains persisted even when the same checkpoints were used to initialize RL (GRPO), improving post-RL ceilings by up to +6.4 compared to RL started from imitation learning (IL).

Efficiency: Less expert data, same optimization budget

A key practical win is demo efficiency. With a fixed optimization budget, Early Experience matches or beats IL using a fraction of expert data. On WebShop, using just 1/8 of the demonstrations with Early Experience already exceeded IL trained on the full demo set; on ALFWorld, parity was hit at 1/2 the demos. The advantage grew with more demonstrations, indicating that agent-generated future states provide supervision signals that demonstrations alone can’t capture.

Where does reinforcement learning fit in?

Early Experience isn’t “RL without rewards”. It’s a supervised recipe that uses agent-experienced outcomes as labels. In environments with verifiable rewards, the team simply adds RL after Early Experience. Because the initialization is better than IL, the same RL schedule climbs higher and faster, with up to +6.4 final success over IL-initialized RL across tested domains. This positions Early Experience as a bridge: reward-free pre-training from consequences, followed (where possible) by standard reinforcement learning (RL).

Key Takeaways

– Reward-free training via agent-generated future states (not rewards) using Implicit World Modeling and Self-Reflection outperforms imitation learning across eight environments.
– Reported absolute gains over IL: +18.4 (WebShop), +15.0 (TravelPlanner), +13.3 (ScienceWorld) under matched budgets and settings.
– Demo efficiency: exceeds IL on WebShop with 1/8 of demonstrations; reaches ALFWorld parity with 1/2—at fixed optimization cost.
– As an initializer, Early Experience boosts subsequent RL (GRPO) endpoints by up to +6.4 versus RL started from IL.
– Validated on multiple backbone families (3B–8B) with consistent in-domain and out-of-domain improvements; positioned as a bridge between imitation learning (IL) and reinforcement learning (RL).

Editorial Comments

Early Experience is a practical and pragmatic contribution. It replaces brittle rationale-only augmentation with outcome-grounded supervision that an agent can generate at scale, without reward functions. The two variants—Implicit World Modeling and Self-Reflection—directly attack off-policy drift and long-horizon error accumulation, explaining the consistent gains over imitation learning across eight environments and the stronger RL ceilings when used as an initializer for GRPO. In web and tool-use settings where verifiable rewards are scarce, this reward-free supervision is the missing middle between IL and RL and is immediately actionable for production agent stacks.

Check out the [PAPER](https://arxiv.org/pdf/2510.08558) here. For tutorials, codes, and notebooks, head over to our [GitHub Page](https://github.com/meta-org/early-experience). You can also follow us on [Twitter](https://twitter.com/metaorg), join our [100k+ ML SubReddit](https://www.reddit.com/r/MachineLearning/), subscribe to our newsletter, and even join us on [Telegram](https://t.me/metaorg).

Hold onto your hats, tech enthusiasts! The Google Pixel 10 Pro Fold, hyped for its durability, has just flopped a major durability test in a spectacular fashion. YouTuber JerryRigEverything put the phone through its paces, and the results are, well, smoking.

First off, the phone snapped along one of its antenna lines while being bent. Now, that’s not great, but it gets worse. As the phone was straightened back out, the battery started smoking so much that it triggered a smoke alarm! 🚨 This is a first in JerryRigEverything’s decade-long history of testing mainstream phones.

Google claims the Pixel 10 Pro Fold can survive up to ten years of folding, but JerryRigEverything wasn’t surprised by the snap, as the previous two foldable Pixels had the same weakness. They described this result as a “catastrophic failure,” likening it to “building a third Death Star with the exact same exhaust port.”

But wait, there’s more! Despite Google’s IP68 claims for dust and water resistance, the Pixel 10 Pro Fold’s hinge can get sand stuck in it. So, it seems only the screen is actually dust-proof, which is not what you’d expect from an IP68 rating.

Now, JerryRigEverything’s tests are extreme, but if you’re planning to buy the Google Pixel 10 Pro Fold based on its durability claims, you might want to be extra careful not to bend it too much when it’s unfolded.

Stay tuned for more tech news and reviews! And if you’re a Google Pixel fan, check out our guide to the best Pixel phones and our review of the Pixel 10 Pro XL.

🚨 Attention, ChatGPT fans! Sam Altman, the mastermind behind OpenAI, has just dropped a bombshell. Get ready for a new version of ChatGPT that’s set to behave more like the beloved GPT-4o model! 🤖💬

Imagine this: customizable tone, personality, and even emoji use! Yes, you heard it right. ChatGPT is about to get a whole lot friendlier and more conversational. No more sounding like a stuffy corporate HR bot! 😎

But wait, there’s more! Altman announced that verified adult users will soon have access to a new “adult mode” that will allow for erotica and other mature content. 🌶️ This is a major policy shift for OpenAI, and it’s sure to spark some serious debate.

So, what’s behind this change of heart? Altman explained that OpenAI has been working hard to improve mental health safeguards and now feels ready to ease up on some restrictions. The goal? To strike a balance between flexibility and safety, creativity and control. 🤝

Now, you might be wondering, what does this mean for you? Well, even if you’re not interested in mature content, you’ll likely notice a difference in how ChatGPT interacts with you. The shift to a more GPT-4o-like personality could be a game-changer. And the best part? It’ll be an opt-in choice, so you can stick with a more straight-laced AI if that’s your preference. 😊

This update is a big deal, folks. It’s a sign that OpenAI is moving away from the one-size-fits-all approach and embracing customization. So, get ready for a ChatGPT that can be as sensitive or sassy, as PG or R-rated as you want it to be. 🎉

Stay tuned for more updates on this exciting development! And if you’re craving more tech news, make sure to follow TechRadar on Google News, TikTok, and WhatsApp! 📱📰🎬

I’m at GITEX Global, Dubai, the world’s biggest tech exhibition, and the buzz is all about AI. But there’s a elephant in the room – energy. Is there enough of it? Can we get it where it’s needed? Octopus Energy’s Co-founder and CEO, Greg Jackson, is here to shed some light, and I caught up with him over coffee just before his talk.

Me: So, Greg, do we have the energy infrastructure needed for AI?

Greg: Not even close. We were struggling before AI, with parts of the UK unable to build houses due to power shortages. Now, AI’s coming in with massive energy demands, and our infrastructure and markets aren’t ready. We’re not even at the starting line.

Me: How so?

Greg: Well, our energy systems are planned years in advance. But AI’s changing everything, and markets can’t respond quickly enough. We need signals to use existing infrastructure better and build new where it’s needed. Plus, we’re wasting energy. In the UK, we’ve spent a billion pounds this year turning off wind farms in Scotland and paying for gas instead. That electricity could power data centers.

Me: Why are we doing that?

Greg: Because the system operators aren’t looking at the grid map. They pay wind farms for electricity, then realize they can’t get it to market due to congestion. So they tell them not to generate, even though they’ve been paid. It’s madness!

Me: How did we get here?

Greg: The world’s changing exponentially, and our systems can’t keep up. Plus, we’ve built generation infrastructure without thinking about consumption. We need economic incentives to build where the energy is. Scotland could have the cheapest electricity in Europe if we did that.

Me: Why is energy so expensive?

Greg: It’s not just the UK. Every state in the US is struggling to build AI infrastructure without hiking energy bills. People need electricity now, but our systems are slow. We could build wind and solar farms in months, but we’re trying to build new nuclear plants that take years.

Me: Can’t we just use batteries?

Greg: Absolutely! Batteries are cheap and getting cheaper. The UK regulator is planning for 2030 using 2019 battery prices. It’s like building landlines after the smartphone was invented. We need to embrace these technologies now.

Me: So, what’s the fix?

Greg: We need more market mechanisms to deploy private capital quickly. Hyperscalers would pay a premium for electricity if they could get it now. We should use that to speed up deployment and bring down costs for everyone. Governments need to loosen their grip on regulations and change them fast to win the AI race.

Me: What can we do in the meantime?

Greg: Governments should use the AI opportunity to change regulations quickly. It would benefit not just AI, but also citizens with cheaper electricity and more generation capacity. It’s a win-win.

And with that, Greg was called away for his talk. I thanked him for his time and the insightful chat. The future of energy and AI certainly looks challenging, but with leaders like Greg pushing for change, there’s hope we can turn this dinosaur system around.