Feed aggregator

When it comes to staying on top of your health, there’s a wealth of technology out there that can give you a boost. The best fitness trackers can help you keep on top of your step count and exercise, while the best smart scales can help you monitor metrics like body composition.

But can you use AI to stay fit and healthy, lose weight, or build muscle? As an experienced health and fitness writer with over a decade using fitness tech like the Apple Watch, I’ve been putting ChatGPT to use as part of my health and fitness regime.

While the usual caveats about AI apply (double-check for hallucinations, etc), I have to say I’ve come away mightily impressed with it as a fitness tool. A lot of the “logistics” of fitness can often involve a lot of minutiae and data, such as calculating macros (proteins, carbs, fat) for your diet. AI is perfect for this sort of work.

So whether you want AI help developing massive workout plans for the weeks and months ahead, a full week’s worth of eating, or more granular help with day-to-day fitness needs, here’s how you can use ChatGPT to help supplement your health and fitness.

Coming up with the right prompts

Fitness and ChatGPT are a lot alike in one key aspect. You really only get out as much as you put in. If you’re half-hearted with your workouts or lax with your diet, your results will reflect that. Likewise, the more precise you are with ChatGPT, and the more effort you put into your prompts, the better the information you’ll get in return.

For example, if you ask ChatGPT, “How do I grow muscle?”, you’ll get some helpful but largely generic advice about progressive overload, reps and sets, and nutrition. None of this is wrong, and if you’re a beginner, ChatGPT can be a handy place to start collating some basic principles of health, exercise, and more.

But I’m here to tell you that ChatGPT is way more powerful than that, and it’s all about going deeper, being more specific, and really stretching the LLM’s legs. As I’ve already hinted at, ChatGPT is useful for large, comprehensive fitness and health tasks, but also for granular day-to-day stuff. Here are some of the prompts I’ve used.

1. Generate a workout plan

(Image credit: OpenAI / Future)

Can you generate an 8-week workout plan that will help me build muscle? I work out in a local gym, so can use most standard equipment, but I only have about 60 minutes a day to exercise.

ChatGPT can be used to generate a full workout plan that can last weeks or even months. This is a great example of the broader, more comprehensive approach to prompts I’m talking about. ChatGPT designed me a six-day workout plan that lasts eight weeks, with an outline of the weekly schedule and a workout breakdown by day.

There are several key components that make the prompt useful. I’ve specified the length of the plan and the goal (eight weeks, build muscle). Crucially, I’ve also specified the level of equipment I have access to, and the length of time I’d like to spend working out.

Each of these is a parameter you can tweak to your own liking, for instance, you could ask for shorter workouts, or specify that you have no equipment so you’ll need to do bodyweight exercises only.

Bonus follow-up: ChatGPT didn’t specify how much rest I should take between exercises, so this is a great follow-up question to ask.

2. Exercise help and lifting form

(Image credit: OpenAI / Future)

Can you tell me about the correct technique for doing a Romanian Deadlift?

Perhaps this is your first time lifting weights, and you’ve spotted an exercise you don’t recognize. You can ask ChatGPT for information and advice about lifting technique, specifying the correct technique to avoid any confusion. A lot of lifting technique is common sense, and I’ve not seen ChatGPT give out shoddy advice in this regard, but ChatGPT even offered me advice on several reputable fitness sources I could check out for video advice if I wanted to get a proper look.

3. Weight and progressive overload

With this workout plan - what sort of weight should I start with, and how do I lift heavier?

If you’re doing any kind of weightlifting or workout plan, especially for the first time, it’s important to know how much weight you should be lifting (and how much you shouldn’t). This is crucial to avoiding injury, but also to ensuring you’re going to make progress over time.

ChatGPT can use a prompt like this to give you general guidelines about progressive overload, but can also be used for specific examples. For instance, you could ask it to give you a progressive overload plan for bench press over a number of weeks.

4. Impromptu training

(Image credit: Future)

I'm on holiday and there's a gym in my hotel, can you generate a quick 30-minute workout for my whole body?

Perhaps you’re not looking for a comprehensive workout plan, but you’re on holiday or a work trip and want to get a quick blast in. A prompt like this can generate a full-body workout in just seconds for you to follow. All you have to specify is your time limits, the target muscle group, and the level of equipment you have access to.

5. Set your diet and nutrient benchmarks

I'm 6 feet tall and weigh 107kg, can you calculate a rough BMR for me, as well as an ideal macronutrient split for fat loss?

If you want to lose weight, pack on muscle, or maintain your physique, knowing your basal metabolic rate (how many calories your body needs each day) is crucial. With just my height and weight, ChatGPT provided a comprehensive picture of my BMR, including estimates of total daily energy expenditure based on activity levels.

Then, it suggested a calorie deficit, and a macronutrient split of protein, carbs, and fats. You can use this information to come up with your own nutrition plan or use this next prompt to get ChatGPT to do it for you.

6. Take the guesswork out of nutrition

(Image credit: Future)

I'd like you to generate a one-week meal plan targeting 2,500 kcal a day. I'd like to eat three times a day, getting 40% of my calories from protein.

If you need a diet or nutrition plan to complement your exercise, you can ask ChatGPT for a nutrition plan specifically tailored to your calorific needs. Just be sure to specify your target calories, how often you’d like to eat, and a general idea of your macronutrient split. You can be more precise too, adding in percentages for carbs and fats.

Bonus followup: ChatGPT can turn the meal plan it has generated for you into a shopping list, organized by food type to make shopping easier.

7. Day-to-day meals

I only have 600 calories left in my intake today, but need to consume around 50 grams of protein. Can you think of a good meal?

If you prefer to eyeball your nutrition or don’t want to follow a daily nutrition plan, you can also ask ChatGPT for on-the-fly meal ideas. I find it helpful to specify the number of calories left that I need, as well as any macronutrient requirements. For this prompt and the previous one, you can also specify any dietary restrictions, intolerances, or just general preferences, too.

Exercise ChatGPT

Each of these prompts is specific and useful, but can also serve as grounding for your one inspiration. For example, just as ChatGPT can create an eight-week lifting program, it can also create a running training schedule, or a plan to help you get better a cycling.

I’ve asked ChatGPT for lifting advice and techniques, but this can also be applied to other disciplines. Likewise, my nutritional goals are largely built around protein and building muscle, but you can tailor your requests to suit your needs, for instance with a focus on fat loss, or on fuelling up properly for endurance work.

Like I said, the key is to apply the principles of fitness to using ChatGPT, the more precise you are and the more effort you make with AI, the more you’ll get out of it.

You might also like...

• We asked AI chatbots Gemini and ChatGPT to design our workouts – then we tried them out
• Personal training is losing its appeal, according to research – and AI apps could be to blame
• I used this AI-powered workout app for an entire year – here’s 5 things I learned

• Here's another sign the Pixel 10 is coming
• It could be the fastest booting Pixel yet
• Expect it to appear sometime in mid-2025

The Google Pixel 9a was made official earlier this week, but we're already well into rumor season for the Pixel 10 due later this year – and the handset just popped up in Android's codebase, with a hint of a useful speed upgrade.

As spotted by Android Authority, the Pixel 10 has been namechecked in the Android Open Source Project (AOSP) code. That's the code on which all the various Android flavors are based, including the Pixel software and Samsung's One UI.

What's more, changes to the boot up process on the Pixel 10 are apparently leading to the handset starting up more quickly, according to a Google engineer – though of course the device will only be in the very early stages of testing right now.

It seems that's down to something called Parallel Module Loading, which basically loads up different bits of Android alongside each other, rather than one at a time – as you might have guessed from the name. The Pixel 10 is showing a 30% speed improvement in this particular part of the boot process, according to code comments.

Camera upgrades rumored

The Pixel 9a ditched the usual Pixel camera bar (Image credit: Google)

This won't be exclusive to the Pixel 10, either: Parallel Module Loading improvements should be rolling out to other recent Android handsets as well. It may well be included as part of Android 16, which is due sometime in the middle of 2025.

As for the Pixel 10, that may well make an appearance in August – which would be a neat 12 months after the launch of the Google Pixel 9 series. To date though, we haven't heard too much in the way of leaks and rumors.

We have seen some leaks in terms of model numbers and internal codenames, but they're not particularly exciting. Of more interest are rumors that the Pixel 10 could come with some camera upgrades, including an extra rear camera.

While the Pixel 9a ditched the iconic rear camera bar that we've grown accustomed to on the Pixel series, it's most probably going to make a return to the Pixel 10 phones. We can expect several different models to make an appearance again, including – perhaps – a successor to the Google Pixel 9 Pro Fold.

You might also like

• Here's why Google is delaying the Pixel 9a
• Older Pixels just got a big performance boost
• The Google Pixel 9a's AI has a RAM problem

Rounding up the big interviews you might have missed, from gardening tips with Martha Stewart to a conversation with a former astronaut about what happens to our bodies after that much time in space.

(Image credit: Spencer Platt)

Rounding up the big interviews you might have missed, from gardening tips with Martha Stewart to a conversation with a former astronaut about what happens to our bodies after that much time in space.

The populist message of Sen. Bernie Sanders' "Fighting Oligarchy" tour has resonated with voters frustrated by President Trump's agenda — and the Democratic Party's muddled response.

(Image credit: Ross D. Franklin)

• SoftBank confirms $6.5 billion deal to acquire Arm-based chipmaker Ampere
• The deal could align Ampere with Arm and previously acquired Graphcore
• The acquisition could put Arm in direct competition with its chip partners

We previously reported that Arm is considering making its own silicon, and that the British firm’s Japanese owner, SoftBank, was considering acquiring Ampere Computing, Arm's only independent server chip vendor, to make this a reality.

Ampere, based in Santa Clara, California and backed by Oracle, designs high-performance, energy-efficient processors optimized for AI and cloud workloads using the Arm compute platform.

An acquisition by SoftBank could shift Arm from simply licensing chip designs to manufacturing its own chips - putting it in direct competition with existing customers and expanding Arm’s footprint in the growing data center space.

Complementing Arm's design strengths

Up until now, the acquisition had only been a strong rumor, but it’s now finally confirmed that SoftBank will be buying Ampere Computing for $6.5 billion (approximately ¥973.0 billion).

The deal, announced on March 19, 2025, will make Ampere a wholly owned subsidiary through SoftBank’s investment arm, Silver Bands 6.

SoftBank Group already owns a majority stake in Arm, and an affiliated company, Arm Technology Investment Ltd, also holds an 8.08% stake in Ampere.

The transaction has already been approved by SBG’s Board of Directors but as always with these things it remains subject to regulatory approvals, including U.S. antitrust clearance and review by the Committee on Foreign Investment in the United States (CFIUS).

In a statement, Softbank said, “Through this strategic alignment following the transaction, Ampere’s expertise in developing and taping out Arm-based chips can be integrated, complementing design strengths of Arm Holdings.”

It added that Ampere is expected to collaborate with “group companies, investees, and business partners.”

SoftBank said the purchase will be financed through borrowings from Mizuho Bank and others. The Raine Group is serving as financial adviser, with Morrison & Foerster providing legal counsel.

Ampere will continue operating under its current structure until the deal closes, which is expected to be in the latter half of 2025.

SoftBank previously acquired UK-based chip designer Graphcore for between $400 million and $500 million.

That company was once considered a potential rival to Nvidia and AMD, but fell on hard times after failing to capitalize on the AI boom.

It's not too much of a stretch to think that Softbank could integrate Ampere with Arm and Graphcore to build a more unified AI compute strategy across its portfolio.

You might also like

• Arm's Japanese owner could buy Arm's only independent server chip vendor
• How Ampere Computing plans to ride the AI wave
• Arm may be planning to hike prices, has considered developing its own chips

Heathrow Airport said it was "fully operational" on Saturday, after an almost daylong closure sparked by an electrical substation fire. But airlines warned that severe disruption will last for days.

(Image credit: Kirsty Wigglesworth)

The advisories come after some citizens from European countries and Canada have been detained and deported by immigration officials while traveling to the United States.

(Image credit: John MacDougall)

Newsmax is being sued for defamation by Dominion Voting Systems — the same company that sued Fox News over false 2020 election-fraud claims and walked away with a nearly $800 million settlement.

(Image credit: BRENDAN SMIALOWSKI/AFP via Getty Images)

As with Europe, the Trump administration is sending conflicting signals to America's long-standing allies in Asia, with whom the U.S. has deep-rooted security agreements that date back to the 1950s.

(Image credit: Handout by South Korean Defense Ministry)

Choose the internet provider with the best plans and widest coverage across Oregon.

This week was full of major announcements. Google showcased the (potentially) iPhone 16e beating Google Pixel 9a, we reviewed Assassin's Creed Shadow and loved it, and Nvidia showcased an actual Star Wars droid in real life.

To catch up on all that and more you can scroll down to see the week's seven biggest tech news stories that you don't want to have missed.

Once you're all up to speed check out our picks for the 7 new movies and TV shows to stream this weekend (March 22).

1. The Google Pixel 9a broke cover

(Image credit: Google)

Google took the covers off the Pixel 9a this week, finally giving us an official look at its latest budget phone. At a quick glance, you’ll probably not notice a huge amount of difference from the Pixel 8a, but look closer, and there are some neat changes.

For starters, the phone’s design has changed with flatter sides and a larger display - which is also brighter - along with a rear camera setup that eschews the pill-shaped design of the Pixel 9 and goes for modules that sit flusher with the phone’s rear.

Under the hood, you’ll find the Google-designed Tensor G4 chip ready to power many Google Gemini activities and generative AI tools. There’s a bigger battery as well. This all comes at the same price as the Pixel 9a’s predecessor, but there’s a slight caveat in that Google has delayed the release of the phone until sometime in April due to a mystery “component quality issue”.

• Read more: Google just launched the Pixel 9a – and I reckon it embarrasses the iPhone 16e

2. We found the droid we've been looking for

Star Wars’ droids just got one step closer to reality at Nvidia's GTC 2025 keynote as Blue joined Jenson Huang on stage to show off the company’s new Newton engine. The robot bounded onto the stage after attendees watched a simulated version of the droid digitally explore the surface of a sandy planet far, far away.

This virtual landscape is what Newton is all about. Rather than needing to actually build a robot and real environments with a wide range of surface designs and materials to train your robot, with Newton, you can instead construct everything digitally, run the simulation, and the AI will gradually learn how to navigate. Once you’ve trained the model virtually, you can upload that data to a real version of the robot, and like Neo training in The Matrix Blue, other bots instantly become experts.

During the keynote, Huang also showcased a bunch of new hardware for developing AI, more humanoid robots, and autonomous vehicle tech.

• Read more: Nvidia GTC 2025 - all the news you might have missed

3. Pebble returned with two new smartwatches

(Image credit: Core Devices)

It’s a big week for retro gadget fans, as Pebble founder Eric Migicovsky’s new Core Devices company unveiled two new smartwatches running the now open-source Pebble OS. The Core 2 Duo is a low-power smartwatch with four buttons, some basic functionalities, an MIP black-and-white screen, and access to Pebble’s suite of 10,000 third-party apps.

The second watch, the Core Time 2, is an upgraded larger version with a color touchscreen and a heart rate monitor. Crucially, both watches last up to 30 days, an impressive 40 times the battery life of your average Apple Watch.

While both are limited runs, Migicovsky said he’ll “make more” if they do well. Available to pre-order now, both devices will be released in July.

• Read more: Pebble founder announces two new smartwatches, and they're basically the opposite of an Apple Watch in every way

4. We reviewed the LG C5 OLED TV

The LG C5 delivers bold, vibrant and dynamic colors, as shown on Disney's Elemental. (Image credit: Future)

We got our hands on one of 2025’s most anticipated TVs, the LG C5 OLED TV. The latest iteration in the ever-popular LG C-series, we spent a week with the C5, and spoiler alert; it’s fantastic.

Earning a full five out of five stars in our review, we loved the LG C5’s stunning picture quality, superb gaming performance, full suite of gaming features, and intuitive smart TV platform with some useful new AI tools. Even its built-in sound was solid!

To address the elephant in the room, no, it’s not that different from its predecessor, the LG C4, and while that’s around, the C5 looks overpriced. But, when the C4’s time is up, the C5 is a fantastic replacement and is already set to be one of the best TVs of 2025.

• Read more: I reviewed the LG C5 OLED TV for over a week, and it's one of the best TVs of the year

5. Deep Research went free to use

(Image credit: Google)

Deep Research is Google Gemini’s AI-powered research assistant. Rather than the usual search and response of the standard chatbot model, using Deep Research, Gemini puts together a full report on whatever you’ve asked it, including citations from sources that it analyzes.

It works best if you throw really meaty questions that need it to consult multiple sources, like “What percentage of dogs in the US are puppies?“, for example. It comes up with a plan for what research it’s going to undertake, which you can edit, and then produces its report. It can take a few minutes for Gemini to assemble a Deep Research report, but when it does, you get a document you can open in Google Docs that covers the subject conclusively.

Best of all, Deep Research is now free to all Gemini users.

• Read more: Gemini Deep Research is now free - here are 4 ways to get the most out of Google’s awesome AI tool

6. Assassin's Creed came out from the Shadows

(Image credit: Ubisoft)

Assassin’s Creed Shadows is out now and is immediately one of our favorites in the entire series. Offering the best open-world role-playing game in that line of Creed games, Shadows gets an awful lot right and not wrong. And I should know, having spent more than 40 hours in the game now.

A major part of the game’s success is the implementation and execution of some of the best combat in the series. Both protagonists, Naoe and Yasuke, offer distinct styles of combat and a unique range of weapons they can use - but each can engage in seriously fluid, chunky, and hard-hitting combat. Whether you’re slashing at enemies with small blades as Naoe, or smashing them about the place with a Kanobo as Yusuke, the combat is Shadows is superb.

Ubisoft Quebec has also rethought world exploration and discovery with Shadows and implemented a much more rewarding and satisfying method of doing so. Instead of viewpoints revealing a sea of known entities in the landscape for you, Shadows only gives you a few points of interest and locations - unmarked, so you have to go and find out what they are. What results is a method of exploring the world that feels rewarding to explore and pulls you on from location to location, landscape to landscape, and region to region.

And while the main story wanes a little, and the Hideout mode that offers cozy base builders a little something can be a drain, there’s just so much to do in Shadows’ Feudal Japan setting that I’m going to be spending dozens more hours in it collecting tea sets, painting wildlife, meditating, learning new combos, carrying out contracts, and assassinating baddies.

• Read more: I’ve played Assassin’s Creed Shadows for more than 40 hours and can’t stop collecting tea sets, painting birds, and assassinating evil schemers

7. Lego and Pokémon broke the internet

(Image credit: Lego / Pokemon)

Lego and Pokémon broke the internet this week when everybody’s favorite plastic bricks announced a partnership with the world’s most valuable media franchise. For years, we’ve hoped for a Pikachu collaboration with Lego, and now our dreams look set to become a reality.

Coming in 2026, perfectly timed with the 30th anniversary of Pokémon, you’ll be able to go on “a real LEGO® Pokémon™ adventure." While we don’t know what LEGO sets will be made available, the teaser trailer showcased Pikachu’s tail, so Pokémon’s mascot is almost a certainty.

We wouldn’t be surprised to see regular releases celebrating the extensive library of Pokémon; after all, there are now over 1,000 species dating back to 1996. With Generation 10 on the horizon and new Nintendo Switch video games set for release this year, it’s definitely an exciting time to be a Pokémon fan.

• Read more: Pokemon and Lego announce the most electrifying collaboration of all time

The Trump administration has announced a flurry of changes at the agency that oversees Social Security. Advocates warn these moves could lead to people having a harder time getting help with benefits.

(Image credit: Kayla Bartkowski)

Here are the answers for The New York Times Mini Crossword for March 22.

• Distributed acoustic sensing detects disturbances in fibre optic signals to identify underwater threats
• NATO’s ‘Baltic Sentry’ mission enhances subsea security, but surveillance remains difficult
• AP Sensing’s North Sea deployment highlights fibre optics’ role in security

Subsea fibre optic cables are a crucial part of global internet infrastructure, yet recent damage incidents in the Baltic Sea have raised concerns about their security.

Per the BBC, there are now efforts to mitigate the risk of sabotage by using a decades-old technique known as distributed acoustic sensing (DAS).

This approach detects disturbances in fibre optic signals by capturing tiny reflections sent back along the strands due to pulses from light encountering vibrations or temperature changes, allowing the system to identify suspicious activity such as underwater drones, vessels dragging anchors, or divers near critical cables.

How fibre optics can 'listen' for threats

As with network security, where businesses rely on the best small business routers to prevent cyber threats, monitoring solutions for subsea infrastructure are becoming essential in safeguarding global communications.

Lane Burdette, a research analyst at TeleGeography, notes that the number of faults affecting subsea cables each year has remained steady, typically between 1 and 200. "Cables break all the time…The number of cable faults per year has really held steady over the last several years."

During tests conducted by AP Sensing, the system detected a diver patting a cable on the seabed, while further experiments demonstrated its ability to identify drones and vessels, potentially providing early warnings of sabotage attempts.

"He stops and just touches the cable lightly, you clearly see the signal...The acoustic energy which travels through the fibre is basically disturbing our signal. We can measure this disturbance," says Daniel Gerwig, global sales manager at AP Sensing, a German technology company.

Just as businesses depend on the best business smartphones for real-time alerts and security updates, early warning systems for subsea cables can provide critical intelligence to prevent disruptions.

Concerns over the vulnerability of these cables have led NATO to launch "Baltic Sentry," a mission using warships, drones, and aircraft to monitor activity in the region, but since constant surveillance is not always possible, demand for fibre optic acoustic sensing solutions is growing.

"It's good that Nato and the European Union have woken up…The question is how quickly you could establish contact with a vessel," said Thorsten Benner, co-founder and director of the Global Public Policy Institute.

Maintaining secure communications in this environment requires the same level of reliability as the best network switches, ensuring smooth data flow and minimal disruption.

Companies such as Optics11 and Viavi Solutions are seeing increased interest in their monitoring technology, which can be deployed on military submarines or along key underwater infrastructure routes.

AP Sensing’s system is already in use in parts of the North Sea, but the technology has limitations, requiring signal interrogation points at regular intervals along the cable and having a sensing range of only a few hundred metres, meaning it can detect nearby threats but is not a complete security solution on its own.

You may also like

• Check out the best cloud storage options around
• Also check out the best cloud backup services
• We’ve curated a list of the best cloud computing services

The heavyweight who lost the "Rumble in the Jungle" to Muhammad Ali before authoring an inspiring second act as a 45-year-old champion and a successful businessman was 76.

(Image credit: Jeff Robbins)

Six Voice of America journalists and a director at its parent agency have sued the Trump administration, alleging its moves to shut down the U.S.-funded network are unconstitutional.

(Image credit: BONNIE CASH/AFP via Getty Images)

• Nvidia’s Quantum-X and Spectrum-X switches reduce reliance on traditional optical transceivers
• Silicon photonics boosts Nvidia’s network efficiency by 3.5x
• Quantum-X and Spectrum-X switches deliver up to 400Tbps throughput for hyperscale AI factories

Nvidia is advancing its networking technology by integrating co-packaged optics (CPO) into its Quantum InfiniBand and Spectrum Ethernet switch, a move expected to reduce power consumption and cost in AI data centers.

At its GTC 2025 event, Nvidia detailed its plans for deploying silicon photonics, which will enhance efficiency by reducing the need for traditional optical transceivers.

Instead of relying on traditional pluggable transceivers, Nvidia is embedding photonics directly into switch ASICs, cutting energy use and minimizing signal loss. These advancements benefit hyperscale AI, and could also improve small business routers with similar efficiency gains.

Nvidia seeks to cut AI data center power by over 50%.

Nvidia’s Spectrum-X and Quantum-X switches use silicon photonics to deliver higher bandwidth and lower energy consumption, supporting up to 1.6 terabits per second (Tbps) per port to efficiently connect millions of GPUs.

The Quantum-X and Spectrum-X photonics switches offer configurations ranging from 128 ports at 800Gbps to 512 ports at 800Gbps, delivering total throughputs of up to 400Tbps.

While the Spectrum-X Ethernet platform enhances multi-tenant hyperscale deployments, the Quantum-X InfiniBand switches deliver superior signal integrity and resilience, making them contenders for the best network switch.

“AI factories are a new class of data centers with extreme scale, and networking infrastructure must be reinvented to keep pace. By integrating silicon photonics directly into switches, NVIDIA is shattering the old limitations of hyperscale and enterprise networks and opening the gate to million-GPU AI factories,” said Jensen Huang, founder and CEO of NVIDIA.

Nvidia’s new switches improve energy efficiency by a factor of 3.5, while also reducing signal degradation. In a typical AI data center with 400,000 GPUs, conventional networking setups require millions of optical transceivers, consuming significant power.

Nvidia’s approach reduces total network power from 72 megawatts to 21.6 megawatts, dramatically improving sustainability. These gains could also enhance business smartphones, enabling faster, more reliable connectivity.

While Nvidia is transitioning towards optical networking, copper remains relevant in specific configurations.

Systems like the GB200 NVL72 still use thousands of copper cables to link GPUs and CPUs via NVLink 5, offering lower power consumption at the rack level.

However, as Nvidia progresses to NVLink 6, copper’s limitations will become more apparent, reinforcing the need for photonic solutions in large-scale AI tool deployments.

Nvidia’s new switches are set for release in late 202 and 2026. The first model, the Quantum 3450-LD InfiniBand switch, launching in late 2025, will provide 144 ports of 800 Gb/sec connectivity and a total bandwidth of 115 Tb/sec.

In 2026, the Spectrum SN6810 Ethernet switch will debut with 128 ports at 800 Gb/sec and an aggregate bandwidth of 102.4 Tb/sec. A larger Spectrum SN6800 model, will also arrive in 2026, featuring 512 ports of 800 Gb/sec and a total throughput of 409.6 Tb/sec.

Via Nextplatform

You may also like

• These are the best hard drives right now
• And here are the best mobile workstations around
• Why you almost certainly have a shadow AI problem

Representatives from Russia and Ukraine will be in meetings to try to hammer out details of a ceasefire on Monday. But peace is still a long way off.

For starters it's only a partial ceasefire—no strikes on energy infrastructure. It's only for 30 days.

And the Ukrainians and Russians aren't even meeting with each other. The U.S. will be a go-between.

One of the biggest things working against a new agreement, is what happened after Ukraine's last agreement with Russia. And the ones before that.

Ukraine says it won't trust a promise from Russia. It needs security guarantees. To understand why, you've got to go back to the birth of independent Ukraine.

For sponsor-free episodes of Consider This, sign up for Consider This+ via Apple Podcasts or at plus.npr.org.

Email us at considerthis@npr.org.

Artificial intelligence needs to get a lot more reliable if it's going to catch up with the hype cycle.