Before the Apple Watch, before the Samsung Galaxy Watch, and before the Google Pixel Watch, there was Pebble. The iconic smartwatch was eventually bought by Fitbit and discontinued in 2016 (and in a cruel twist, the same fate befell many of the best Fitbits after Fitbit was bought by Google), but many users still have fond memories of their Pebbles, and an active community of users are still maintaining functional Pebbles.
Now, it seems like Pebble is ready for a comeback. Google, with no plans to disrupt sales of the Google Pixel Watch 3 and Fitbit’s best fitness trackers, don’t seem to care to release an official Pebble, but they are throwing a bone to the community by preparing to make PebbleOS open source, according to a recent blog post.
For the uninitiated, open source means that anyone will be able to download the source code for Pebble watches, make changes, and publish their versions to the internet for free. This will allow tech-savvy Pebble fans to keep their watches operational - and pave a way for the iconic smartwatch’s return under a new name.
Eric Migicovsky, Pebble’s founder, told TechCrunch that he’s devoted himself full time to creating a new third-party smartwatch. In a blog post on his own website, he writes; “The company behind it failed but millions of Pebbles in the world kept going, many of them still to this day.
(Image credit: Future / Lance Ulanoff)“I wear my Pebble every day. It's been great (and I'm astounded it’s lasted 10 years!), but the time has come for new hardware.”
Migicovsky is the project’s only current full-time employee. Using PebbleOS, the future smartwatch, which will be released under a different name than Pebble, will (according to Migicovsky’s blog post) conform to the following tenets:
You can sign up for information on the project at rePebble.com.
Analysis: Migicovsky’s perfect watchMigicovsky’s ideal watch, on the surface, certainly sounds like one I’d wear in rotation. I love the memory-in-pixel screen you can get on the best Garmin watches and other digital watch-style aesthetics, so the e-paper screen with minimal distractions, long battery life and physical buttons are music to my ears.
With watches reaching a sort of inflection point, I believe we’ve already hit ‘peak smartwatch’ and future models are going to look more and more similar from hereon out. A back-to-basics, open source approach sounds like the breath of fresh air this form factor needs. What’s old is new again.
However, I’m acutely aware the project isn’t really off the ground yet, and may never come to fruition. The smartwatch industry has moved on from Pebble, and as much as some users might see it as a welcome change, there’s every chance most users will be too comfortable with their pre-existing watches to give an upstart any time of day. We’ll continue reporting on the new Pebble as new information comes to light.
You might also like...Trump said the diversity programs discourage merit and leadership, and discriminate on the basis of race and sex. The directive was among a series of orders that seeks to reshape the U.S. military.
(Image credit: Anna Moneymaker)
As 2025 arrives, it’s clear that the internet’s vast capabilities – spanning cloud services to emerging technologies like AI – depend on robust infrastructure. From smart cities to streaming services, while data might be the lifeblood of modern organizations, connectivity is the beating heart. No single technology can meet the world’s growing and diverse connectivity demands – nor should it. Instead, an intricate digital tapestry will emerge, reshaping industries and transforming the global economy.
To paint a picture of this future, here are my predictions for connectivity in 2025.
1. More people will move beyond fiber – embracing the mmWave spectrumThe notion that fiber will reach every corner of the world will be exposed as unrealistic and undesirable. While fiber makes sense in many areas, it’s not a silver bullet; financial and logistical challenges make it impractical in low-density regions. As such, mmWave will gain more recognition as a practical and desirable complement to fiber. Offering gigabit speeds wirelessly in hard-to-reach areas, mmWave delivers reliable, high-speed internet without the extensive groundwork required for fiber installation – and the wheels are already in motion.
Recently, Verizon unveiled an ambitious plan to double its fixed wireless access (FWA) subscribers to 8-9 million by 2028. By deploying mmWave radio frequency (RF) technology, the operator is targeting coverage to 90 million households, having reached its goal of 4-5 million 15 months early. This momentum reflects the growing importance of mmWave in delivering high-speed broadband services.
In 2025, expect to see a surge in the adoption of mmWave technology, particularly in markets where traditional fiber rollouts are impractical or costly. For example, the UK is preparing for the mmWave spectrum auction, unlocking high-frequency bands that promise blazing-fast 5G and transformative services across industries. From consumer electronics to smart cities, 2025 will reveal just how critical mmWave bands are for transforming our digital economy.
2. Smart cities will adopt wireless infrastructure as a key complement to fiberWell-chosen technology will make cities safer while improving accessibility and sustainability. Local government leaders that invest in high-capacity, low-latency technologies will be able to support top-tier CCTV cameras, sensors, autonomous systems, and smart grids – strengthening security and improving outcomes for citizens. Those that don’t will struggle with inefficiencies, limited scalability, and safety risks.
For example, 4K CCTV cameras will resolve its most common problem: that CCTV is too low quality to prevent crime effectively or be used convincingly as evidence. Wireless outdoor infrastructure will enable widespread 4K CCTV and leave this problem in the past where it belongs.
Also on the rise are smart poles with integrated connectivity options, which provide urban areas with efficient and scalable networking solutions. According to ABI Research, more than 10.8 million smart poles will be installed by 2030, and it’s expected that 20% will need wireless connectivity. Every pole will need to deliver gigabits per second; as a critical connectivity option for capacity-hungry applications, mmWave will prove a key enabler. With leading players working actively on new solutions, next year will see this shift continue.
And let’s not forget autonomous vehicles: this might seem like a faraway future, but 2024 saw Waymo take driverless technology to new heights with its robotaxis, so it’s only a matter of time before this becomes a reality for public transport. Once this happens, smart vehicles like buses and trams will generate vast amounts of data that needs to be transferred to and from the cloud. This won’t be possible without wireless connectivity; city leaders that prioritize this when IT infrastructure planning will unlock not only greater efficiency but also adaptability, enabling them to keep up with constantly evolving demands.
3. mmWave-based fixed wireless access will help bridge the digital divideIn developed regions and emerging countries alike, the digital divide remains one of the most pressing connectivity challenges worldwide. In established regions like North America, underserved rural areas often lack reliable high-speed internet, despite significant infrastructure investments elsewhere in the country. Meanwhile, in emerging markets like Africa, even cities face limited internet access due to a lack of wireline connectivity and congestion of traditional spectrum options.
2025 will see mmWave-based FWA materialise as a powerful solution to plug these gaps. This is because mmWave technologies operating in the 60GHz band offer a cost-effective way to deliver ultrafast, low-latency connectivity. By bypassing the need for extensive physical infrastructure, mmWave will help democratize internet access and unlock new economic potential on a global scale.
4. 60GHz will gain traction in wireless video and enterprise applicationsIn 2025, the 60GHz spectrum will come into its own, transforming wireless video applications across sectors. From immersive gaming and entertainment experiences to enterprise-grade video conferencing, ultrafast and low-latency connectivity will underpin the rise of professional applications.
Virtual reality (VR), wireless HDMI, and ultra-wide screens will increasingly rely on the speed and quality of 60GHz connectivity. These developments will also play a crucial role in facilitating impactful AI and big data analysis, alongside cloudification and network function virtualization (NFV).
Additionally, as consumer and enterprise demands grow, it will become increasingly understood that business success depends not just on talent and innovation but on the speed and quality of supporting connectivity infrastructure. As 2025 will prove, time will only become a more precious commodity, so anything that speeds up progress reliably will be most welcome.
So what does this mean for infrastructure suppliers? Pushing the boundaries of connectivity — from power efficiency to manufacturability — means tackling competing demands for high performance, reliability, and energy efficiency, while striving to make products accessible and affordable. It's a constant balancing act between advancing innovation and meeting the real needs of global connectivity; striking that balance will shape 2025 as much as any technical milestone. In practice, this means adopting an adaptable approach that doesn’t just solve today’s problems but also anticipates those of tomorrow. On behalf of all digital citizens, current and future: bring on 2025.
We've compiled a list of the best cloud storage.
This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
As Ukraine's line of defense grows thin, this unit is using a modernized Soviet-era vehicle to stop Russian forces from crossing the river and taking Kherson.
(Image credit: Anton Shtuka for NPR)
Historians say the Trump-ordered release of more information on the killings of President John F. Kennedy, Robert F. Kennedy and Martin Luther King Jr., could be interesting but unlikely to rewrite history.
(Image credit: AFP)
An exhibition at San Francisco's Asian Art Museum points to a burgeoning trend: museums are engaging the public more openly around efforts to repatriate artifacts looted from other countries.
(Image credit: Kevin Candland)
Denmark's last minister for Greenland says Britain once claimed right of first refusal over the territory — because of its proximity to Canada, which Britain once ruled.
(Image credit: Felipe Dana)
The role of first lady is unelected, unpaid, and usually comes with zero preparation. But Melania Trump is back in the White House with the most experience someone in her position can have.
(Image credit: Greg Nash)
Many health professionals are lining up against Trump's pick for health secretary. They say his anti-vaccine views could cost lives. Some of his supporters embrace his stance.
(Image credit: Anna Moneymaker)
Denmark's prime minister went on a tour of European capitals Tuesday as Europe faces what she called "a more uncertain reality" and her country moves to strengthen its military around Greenland.
(Image credit: Antti Aimo-Koivisto)
The software development landscape of 2024 has revealed both the potential and limitations of current AI coding tools. While 74% of developers have embraced these tools, a revealing Stack Overflow study shows 36% struggling with code reliability. This isn't a failure of AI – it's a clear indication of where AI tools need to evolve.
The emergence of "AI Debt" – the hidden costs arising from hastily deployed AI-generated code – in industry discussions highlights a crucial challenge: AI-generated code often requires extensive review and optimization before it's production-ready, diminishing the efficiency gains made in the initial code generation process.
However, studies showing tools like Copilot introducing up to 41% more bugs point not to AI's limitations, but to the need for more sophisticated approaches to AI-assisted development. What are these approaches?
Looking ahead to 2025, here are five key developments that will transform how we use AI in software development.
1. Intelligent Context Modelling Will Transform Code GenerationNext-generation AI will move beyond simple pattern matching to true contextual understanding. These systems will build comprehensive models of your codebase, architecture, and development patterns, ensuring every suggestion fits seamlessly into your existing ecosystem.
So instead of producing isolated snippets of code, they will provide suggestions that align with the broader software design and help to predict issues in performance, security and scalability.
This deep context awareness will dramatically reduce the current effort required to adapt AI-generated code to production environments. And not only will this save on developer time, but contribute to a new standard in quality for creating performance-ready AI-generated code.
2. Multi-Large Language Model (LLM) Architectures Will Replace Single-Model DependenciesThe limitations of single-model approaches will give way to sophisticated multi-LLM architectures. These systems will treat coding LLMs as modular infrastructure components, using advanced prompt engineering and model orchestration to leverage the strengths of different models. For example, one model may be optimized for code syntax and another for code refactoring.
This means companies can access parallel processing and use different LLMs best suited to process different tasks, while also enhancing their reliability by being less dependent on one model. Such adaptability will also allow companies to scale more effectively and cost-efficiently.
This architectural shift will free developers from vendor lock-in while enabling more sophisticated code generation and optimization capabilities. We’re already seeing this start to happen as the benefit of multi-modal becomes more apparent.
3. AI-Driven Code Evolution Through Genetic AlgorithmsStatic code generation will evolve into dynamic code optimization through genetic algorithms. Based on Darwinian principles, these systems will continuously generate, test, and refine code variations, automatically selecting the best performers based on specific metrics. For instance, the first batch of AI-generated code is the gene pool, which then, through the use of genetic algorithms, undergoes evolutionary processes, with code tested against metrics such as processing efficiency and memory usage. It’s survival of the fittest but for code optimization.
This evolutionary approach ensures code continuously improves as system and business requirements change rather than remaining static after initial generation.
4. Automated Validation Will Shift Left in DevelopmentReal-time validation will become an embedded part of the development process. AI systems will automatically verify security, performance, and compatibility as code is written, not after, ensuring low-quality code is filtered out during the process.
This shift-left approach will integrate comprehensive testing and validation directly into the development workflow, significantly reducing post-generation review time. Ultimately, this will accelerate the overall software development cycle while also improving quality.
5. Next-Gen Intelligence Platforms Will Revolutionize Developer WorkflowsThe culmination of these advances will be intelligent platforms that fundamentally transform development workflows. These platforms will orchestrate multiple AI technologies while continuously learning from every interaction, code review, and deployment outcome.
What makes these platforms revolutionary is their ability to evolve alongside your development practices. By learning from successful implementations, failed attempts, and developer feedback, they'll become increasingly sophisticated in their understanding of what makes code not just functional, but optimal for specific contexts and requirements. They will evolve and improve with each iteration.
For developers, this evolution means moving beyond simple code completion to truly intelligent development assistance that understands your unique technical environment and objectives. These platforms won't just suggest code – they'll help create better, more reliable software while reducing the manual overhead that currently limits AI's potential in development.
Making the move to intelligent development platformsThe rapid development and hype around AI has led to a majority of software developers adopting AI tools for coding. But with this widespread adoption, the next step in the evolution of these tools is to significantly improve the reliability, quality and performance of AI-generated code. With current processes, much skill, time and effort is required to adapt and maintain code after it is generated. This is where the next-generation of tools will start to make their impact.
Intelligent context modelling and multi-LLM architectures will be a new breed of tools significantly reducing the effort involved in code generation and enhancing optimization capabilities. When code is being generated, genetic algorithms will use natural selection principles to ensure the best lines of code remain, while real-time validation will play its part in enhancing quality as the code is being written.
These advances will culminate in the next generation of intelligent platforms which continuously learn and evolve alongside a developer’s specific practices. Ultimately, in 2025, rather than simply using code generation tools, developers will begin to transform their processes with truly intelligent AI assistants.
We've compiled a list of the best laptops for programming.
This article was produced as part of TechRadarPro's Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/news/submit-your-story-to-techradar-pro
A message to USAID staff, obtained by NPR, says it will analyze "actions within USAID that appear to be designed to circumvent the President's Executive Order" freezing most foreign aid.
(Image credit: Celal Gunes/Anadolu Agency via Getty Images)
Meta AI wants to be more like your good friend who remembers all the important details about you and less like an impersonal AI chatbot. Meta is rolling out updates to its AI assistant across Messenger, WhatsApp, Facebook, and Instagram in the U.S. and Canada to do just that.
Those engaging with Meta AI may find it remembers much more about you between conversations, including some details culled from your profiles on the various connected platforms that can personalize its recommendations.
The new Memory Boost feature is pretty much exactly what it sounds like. Meta AI can now remember key details you share in one-on-one chats. Maybe you casually mention that you love hiking or that you’re fluent in Italian. Meta AI will store that tidbit in its digital brain and use it to inform future responses. A friend who remembers how you like your coffee is a good friend. Whether that same is true about an AI pal is debatable. Especially as, in this case, it’s a friend who is also keeping track of your brunch preferences, weekend plans, and which country artists you’ve been streaming.
The system is also context-sensitive, which means Meta AI doesn’t need you to spoon-feed it all the details. If you tell it you’re vegan while asking for breakfast ideas, it’ll note that for the future and stop recommending omelets or fancy frittatas. This memory feature only applies to one-on-one chats, not group conversations; you can also delete its memory whenever you want.
Personal geographyMeta AI won't just remember your personality quirks. AI is becoming more personalized by tapping the data you share via Meta's platforms. For instance, if you’re looking for family activities this weekend, Meta AI might suggest a local concert based on your location, your recent browsing habits, and its memory of you mentioning that you have two kids.
Of course, this level of personalization isn’t entirely new. Meta has been perfecting the art of AI-driven recommendations for years. What makes this different is the assistant’s ability to combine information from multiple sources, including your chat history, profile, and viewing habits, to deliver something relevant.
Trying to stand out by offering this much personalization makes sense as Meta tries to break through the clamor around AI assistants like ChatGPT, Google Gemini, and others. Meta’s strategy seems to be doing what it can to embed Meta AI within its existing ecosystem of platforms used by millions of people every day. After all, why would you open a separate AI app when you can just ask Meta AI directly on WhatsApp or Facebook?
You might also likeAyaneo has been putting out a wide range of high-quality handheld gaming devices. While the Chinese hardware company mainly focused on Android-based portables in 2024, it’s returning to form with the new Windows-based flagship handheld first revealed in November.
The Ayaneo 3 is a direct sequel to its Ayaneo 2, which launched in February 2023 and was released as a rival to the Steam Deck. Not only does the Ayaneo 3 upgrade the specs, but it also adds new features to enhance the gaming experience.
In terms of specs, there are two models to choose from. The first comes equipped with an AMD Ryzen 7 8840U CPU with eight cores and 16 threads, 3.3GHz clock speed, and an NPU of 16 TOPS.
The second model features the AMD Ryzen AI 9 HX 370 CPU with 12 cores and 24 threads, a base clock speed of 2GHz, and an NPU at 50TOPS. The former will be available in every configuration except the 4TB RAM one, while the latter starts at 32GB instead of 16GB and goes from there.
You can also choose between 7-inch LCD and OED displays for any model with different displays. The LCD screen has a refresh rate of 120Hz/60Hz and maxes out at 500 nits. The OLED has a 144Hz/120Hz/90Hz/60Hz refresh and maxes out at 800 nits.
Why its modular controls matterWhat really sets the Ayaneo 3 apart from the rest is its modular controls, which are unique to pretty much any other gaming handheld. Modular controls mean the directional pad, analog stick, and buttons can be physically switched around or rotated to give a more customizable experience. This lets you match controls to controllers like Playstation and Xbox or create your own fighting stick for titles like Street Fighter 6.
The controls themselves are also decked out spec-wise. The joysticks and triggers use Hall-effect sensors with a 4ms response time. The triggers also have two modes: linear mode, which gives you 7.5mm of travel distance to enable precise analog input, and tactile mode, which reduces the travel to 3mm for the repeated button presses known in FPS games.
Right now, you can support the Ayaneo 3 in its crowdfunding campaign on Indiegogo. The price ranges from $700 for the 16GB/512GB model without a complete set of input modules and $1,800 for the 64GB/4TB with a full set. An extra set runs $100, while a backup storage case — in Starry Black, Sky White, and Retro Power — costs $33.
Though the upper pricing is an eye-watering amount (and the lowest amount is still nothing to sneeze at), it’s meant as a premium portable system with unique features. Its selling point is that its competition, like the Steam Deck or Lenovo Legion Go, can’t match it in that sense. And if the Ayaneo 3 finds its footing, it could carve out a whole new niche in the gaming handheld market.
OpenAI has enhanced ChatGPT's collaborative editing feature, Canvas, with some major upgrades. The updates make Canvas even more flexible in helping you compose and edit content and code. The brainpower boost also comes with better visuals and some love for macOS desktop users.
The most significant upgrade is that Canvas will now work with OpenAI’s new o1 model. The more deliberative o1 model is already better than previous iterations at complex reasoning and problem-solving. With o1, Canvas is better at collaborating and explaining its logic.
Imagine it more like a human partner and less like a hyper-efficient autocorrect. You can enable the o1 model in Canvas by selecting it from the model picker or typing the command “/canvas.” Free users are out of luck for now, though, as only Pro, Plus, and Team users can access the o1 model in Canvas.
But it’s not just brains; Canvas has also received a visual boost – it also now supports rendering React and HTML code. That means you can actually see your projects come to life right in the workspace. You don't have to imagine how the code would perform in a browser or run a separate testing system. Better yet, free ChatGPT users can play with this new ability right now.
Regardless of subscription level, Mac users keen to use ChatGPT more are also getting more access to the AI chatbot feature. There's now full Canvas support on the macOS ChatGPT desktop app, allowing you to do anything Canvas can provide without leaving it.
Canvas companionThe o1 model’s more deliberate thinking aims to address criticisms that AI often feels rushed or too eager to please. With the added ability to visualize React and HTML code, OpenAI has made it clear that it wants Canvas to be a playground for developers, not just a chatroom with a few extra bells and whistles.
That fits with OpenAI's and other AI developers' keenness to make AI chatbots and tools more dynamic. Editing directly within the Canvas interface and being able to comment on and prompt changes to just part of an AI answer is a subtle but powerful improvement over the standard back-and-forth.
Just being able to drop text into Canvas and assigning the AI to advise or help edit is no small thing. Sometimes, you just want criticism without it being redone. That goes double if you share complex code or a very long text.
Ultimatley, Making ChatGPT more proactive and adaptable when it comes to carrying out very specific requests is likely to be a big part of how AI is employed in the future.
You might also like