Anthropic on Tuesday released Claude Sonnet 4.6, a model that amounts to a seismic repricing event for the AI industry. It delivers near-flagship intelligence at mid-tier cost, and it lands squarely in the middle of an unprecedented corporate rush to deploy AI agents and automated coding tools.
The model is a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. It features a 1M token context window in beta. It is now the default model in claude.ai and Claude Cowork, and pricing holds steady at $3/$15 per million tokens — the same as its predecessor, Sonnet 4.5.
That pricing detail is the headline that matters most. Anthropic’s flagship Opus models cost $15/$75 per million tokens — five times the Sonnet price. Yet performance that would have previously required reaching for an Opus-class model — including on real-world, economically valuable office tasks — is now available with Sonnet 4.6. For the thousands of enterprises now deploying AI agents that make millions of API calls per day, that math changes everything.
Anthropic’s computer use scores have nearly quintupled in 16 months. The company’s latest model, Sonnet 4.6, scored 72.5 percent on the OSWorld-Verified benchmark, up from 14.9 percent when the capability first launched in October 2024. (Source: Anthropic)
Advertisement
Why the cost of running AI agents at scale just dropped dramatically
To understand the significance of this release, you need to understand the moment it arrives in. The past year has been dominated by the twin phenomena of “vibe coding” and agentic AI. Claude Code — Anthropic’s developer-facing terminal tool — has become a cultural force in Silicon Valley, with engineers building entire applications through natural-language conversation. The New York Times profiled its meteoric rise in January. The Verge recently declared that Claude Code is having a genuine “moment.” OpenAI, meanwhile, has been waging its own offensive with Codex desktop applications and faster inference chips.
The result is an industry where AI models are no longer evaluated in isolation. They are evaluated as the engines inside autonomous agents — systems that run for hours, make thousands of tool calls, write and execute code, navigate browsers, and interact with enterprise software. Every dollar spent per million tokens gets multiplied across those thousands of calls. At scale, the difference between $15 and $3 per million input tokens is not incremental. It is transformational.
The benchmark table Anthropic released paints a striking picture. On SWE-bench Verified, the industry-standard test for real-world software coding, Sonnet 4.6 scored 79.6% — nearly matching Opus 4.6’s 80.8%. On agentic computer use (OSWorld-Verified), Sonnet 4.6 scored 72.5%, essentially tied with Opus 4.6’s 72.7%. On office tasks (GDPval-AA Elo), Sonnet 4.6 actually scored 1633, surpassing Opus 4.6’s 1606. On agentic financial analysis, Sonnet 4.6 hit 63.3%, beating every model in the comparison, including Opus 4.6 at 60.1%.
These are not marginal differences. In many of the categories enterprises care about most, Sonnet 4.6 matches or beats models that cost five times as much to run. An enterprise running an AI agent that processes 10 million tokens per day was previously forced to choose between inferior results at lower cost or superior results at rapidly scaling expense. Sonnet 4.6 largely eliminates that trade-off.
Advertisement
In Claude Code, early testing found that users preferred Sonnet 4.6 over Sonnet 4.5 roughly 70% of the time. Users even preferred Sonnet 4.6 to Opus 4.5, Anthropic’s frontier model from November, 59% of the time. They rated Sonnet 4.6 as significantly less prone to over-engineering and “laziness,” and meaningfully better at instruction following. They reported fewer false claims of success, fewer hallucinations, and more consistent follow-through on multi-step tasks.
Anthropic’s Sonnet 4.6, a mid-tier model, matches or approaches the performance of the company’s flagship Opus line across most benchmark categories — and frequently outperforms rival models from Google and OpenAI. (Source: Anthropic)
How Claude’s computer use abilities went from ‘experimental’ to near-human in 16 months
One of the most dramatic storylines in the release is Anthropic’s progress on computer use — the ability of an AI to operate a computer the way a human does, clicking a mouse, typing on a keyboard, and navigating software that lacks modern APIs.
When Anthropic first introduced this capability in October 2024, the company acknowledged it was “still experimental — at times cumbersome and error-prone.” The numbers since then tell a remarkable story: on OSWorld, Claude Sonnet 3.5 scored 14.9% in October 2024. Sonnet 3.7 reached 28.0% in February 2025. Sonnet 4 hit 42.2% by June. Sonnet 4.5 climbed to 61.4% in October. Now Sonnet 4.6 has reached 72.5% — nearly a fivefold improvement in 16 months.
Advertisement
This matters because computer use is the capability that unlocks the broadest set of enterprise applications for AI agents. Almost every organization has legacy software — insurance portals, government databases, ERP systems, hospital scheduling tools — that was built before APIs existed. A model that can simply look at a screen and interact with it opens all of these to automation without building bespoke connectors.
Jamie Cuffe, CEO of Pace, said Sonnet 4.6 hit 94% on their complex insurance computer use benchmark, the highest of any Claude model tested. “It reasons through failures and self-corrects in ways we haven’t seen before,” Cuffe said in a statement sent to VentureBeat. Will Harvey, co-founder of Convey, called it “a clear improvement over anything else we’ve tested in our evals.”
The safety dimension of computer use also got attention. Anthropic noted that computer use poses prompt injection risks — malicious actors hiding instructions on websites to hijack the model — and said its evaluations show Sonnet 4.6 is a major improvement over Sonnet 4.5 in resisting such attacks. For enterprises deploying agents that browse the web and interact with external systems, that hardening is not optional.
Enterprise customers say the model closes the gap between Sonnet and Opus pricing tiers
The customer reaction has been unusually specific about cost-performance dynamics. Multiple early testers explicitly described Sonnet 4.6 as eliminating the need to reach for the more expensive Opus tier.
Advertisement
Caitlin Colgrove, CTO of Hex Technologies, said the company is moving the majority of its traffic to Sonnet 4.6, noting that with adaptive thinking and high effort, “we see Opus-level performance on all but our hardest analytical tasks with a more efficient and flexible profile. At Sonnet pricing, it’s an easy call for our workloads.”
Ben Kus, CTO of Box, said the model outperformed Sonnet 4.5 in heavy reasoning Q&A by 15 percentage points across real enterprise documents. Michele Catasta, President of Replit, called the performance-to-cost ratio “extraordinary.” Ryan Wiggins of Mercury Banking put it more bluntly: “Claude Sonnet 4.6 is faster, cheaper, and more likely to nail things on the first try. That combination was a surprising combination of improvements, and we didn’t expect to see it at this price point.”
The coding improvements resonate particularly given Claude Code’s dominance in the developer tools market. David Loker, VP of AI at CodeRabbit, said the model “punches way above its weight class for the vast majority of real-world PRs.” Leo Tchourakov of Factory AI said the team is “transitioning our Sonnet traffic over to this model.” GitHub’s VP of Product, Joe Binder, confirmed the model is “already excelling at complex code fixes, especially when searching across large codebases is essential.”
Brendan Falk, Founder and CEO of Hercules, went further: “Claude Sonnet 4.6 is the best model we have seen to date. It has Opus 4.6 level accuracy, instruction following, and UI, all for a meaningfully lower cost.”
Advertisement
In a simulated business environment, Sonnet 4.6 nearly tripled the earnings of its predecessor over the course of a year, suggesting sharply improved decision-making in complex, long-horizon tasks. (Source: Anthropic, Vending-Bench Arena)
A simulated business competition reveals how AI agents plan over months, not minutes
Buried in the technical details is a capability that hints at where autonomous AI agents are heading. Sonnet 4.6’s 1M token context window can hold entire codebases, lengthy contracts, or dozens of research papers in a single request. Anthropic says the model reasons effectively across all that context — a claim the company demonstrated through an unusual evaluation.
The Vending-Bench Arena tests how well a model can run a simulated business over time, with different AI models competing against each other for the biggest profits. Without human prompting, Sonnet 4.6 developed a novel strategy: it invested heavily in capacity for the first ten simulated months, spending significantly more than its competitors, and then pivoted sharply to focus on profitability in the final stretch. The model ended its 365-day simulation at approximately $5,700 in balance, compared to Sonnet 4.5’s roughly $2,100.
This kind of multi-month strategic planning, executed autonomously, represents a qualitatively different capability than answering questions or generating code snippets. It is the type of long-horizon reasoning that makes AI agents viable for real business operations — and it helps explain why Anthropic is positioning Sonnet 4.6 not just as a chatbot upgrade, but as the engine for a new generation of autonomous systems.
Advertisement
Anthropic’s Sonnet 4.6 arrives as the company expands into enterprise markets and defense
This release does not arrive in a vacuum. Anthropic is in the middle of the most consequential stretch in its history, and the competitive landscape is intensifying on every front.
On the same day as this launch, TechCrunch reported that Indian IT giant Infosys announced a partnership with Anthropic to build enterprise-grade AI agents, integrating Claude models into Infosys’s Topaz AI platform for banking, telecoms, and manufacturing. Anthropic CEO Dario Amodei told TechCrunch there is “a big gap between an AI model that works in a demo and one that works in a regulated industry,” and that Infosys helps bridge it. TechCrunch also reported that Anthropic opened its first India office in Bengaluru, and that India now accounts for about 6% of global Claude usage, second only to the U.S. The company, which CNBC reported is valued at $183 billion, has been expanding its enterprise footprint rapidly.
Meanwhile, Anthropic president Daniela Amodei told ABC News last week that AI would make humanities majors “more important than ever,” arguing that critical thinking skills would become more valuable as large language models master technical work. It is the kind of statement a company makes when it believes its technology is about to reshape entire categories of white-collar employment.
The competitive picture for Sonnet 4.6 is also notable. The model outperforms Google’s Gemini 3 Pro and OpenAI’s GPT-5.2 on multiple benchmarks. GPT-5.2 trails on agentic computer use (38.2% vs. 72.5%), agentic search (77.9% vs. 74.7% for Sonnet 4.6’s non-Pro score), and agentic financial analysis (59.0% vs. 63.3%). Gemini 3 Pro shows competitive performance on visual reasoning and multilingual benchmarks, but falls behind on the agentic categories where enterprise investment is surging.
Advertisement
The broader takeaway may not be about any single model. It is about what happens when Opus-class intelligence becomes available for a few dollars per million tokens rather than a few tens of dollars. Companies that were cautiously piloting AI agents with small deployments now face a fundamentally different cost calculus. The agents that were too expensive to run continuously in January are suddenly affordable in February.
Claude Sonnet 4.6 is available now on all Claude plans, Claude Cowork, Claude Code, the API, and all major cloud platforms. Anthropic has also upgraded its free tier to Sonnet 4.6 by default. Developers can access it immediately using claude-sonnet-4-6 via the Claude API.
Platinum-group metals (PGMs) are great catalysts, but they’re also great investments — in the sense that they are very, very expensive. Just ask the guy nicking car exhausts in the Walmart parking lot. If one could replace PGMs with a more common element, like, say the aluminum that makes up over 8% the mass of this planet, it would be a boon to the chemical industry, and a bane to meth addicts. Researchers at King’s College, London have found a way to do just that, with a novel form of aluminum called cyclotrialumane.
The aluminum trimer is exactly what the ‘tri’ in the name makes it sound like: three aluminum atoms, bonded in a triangular structure that is just pointy and stick-outy enough to poke into other molecules and make chemistry happen. OK, not really — you can see from the diagram above it’s not nearly that simple — but the point is that the shape makes it a good catalyst. The trimer structure is useful in large part because it is very stable, allowing reactions to be catalyzed in a large variety of solutions.
The researchers specifically call out their trialuminum compound as effective at splitting H2 in to H+ ions, as well as ethene polymerization. Both of those are important industrial reactions, but that’s only a start for this trialuminum wonder catalyst, because the researchers claim it can catalyze totally new reactions and create previously-unknown chemicals.
If you never took chemistry, or it’s been too many years since you last slept through that class, we have a primer on catalysts here. By accelerating chemical reactions, catalysts have enabled some neat hacks, like anything involving platinum-cure silicone.
Advertisement
Thanks to [Lightislight] for the tip! Hacks do appear here on their own, but you can always use our tips line to catalyze the synthesis of a particular article.
Header image adapted from: Squire, I., de Vere-Tucker, M., Tritto, M. et al. A neutral cyclic aluminium (I) trimer. Nat Commun17, 1732 (2026). https://doi.org/10.1038/s41467-026-68432-1
Some days, it feels like we’re getting all the bad parts of cyberpunk and none of the cool stuff. Megacorps and cyber warfare? Check. Flying cars and holograms? Not quite yet. This week, things took a further turn for the dystopian with the news that a woman was hospitalized after an altercation with a humanoid robot in Macau. Police arrived on scene, took the bot into custody, and later told the media they believed this was the first time Chinese authorities had been called to intervene between a robot and a human.
The woman, reportedly in her seventies, was apparently shocked when she realized the robot was standing behind her. After the dust settled, the police determined it was being operated remotely as part of a promotion for a local business. We’ve heard there’s no such thing as bad publicity, but we’re not sure the maxim holds true when you manage to put an old lady into the hospital with your ad campaign.
Speaking of robots, the U.S. Library of Congress recently discovered and subsequently restored Georges Méliès’s Gugusse et l’Automate (Gugusse and the Automaton), a short film from 1897 that’s considered the first piece of science fiction cinema. As far as anyone knows, it’s also the first time a robot appeared on screen, although this isn’t exactly The Terminator we’re talking about here.
The runtime is less than a minute, but to make the short story even shorter: a guy cranks up a robot that gets bigger and bigger until it turns on its maker and starts to hit him with a stick. The human responds in kind by smashing the robot with a cartoonishly large mallet until it poofs out of existence. The modern film school interpretation is that it’s a cautionary tale about the dangers of technology, ye old Black Mirror, if you will. Since nobody can ask old Georgie what he was going for, we’ll just have to take their word for it.
Advertisement
Returning to the desert of the present, Tom’s Hardware reports that at least one manufacturer is starting to pack their new RAM with an additional non-functioning filler module. With prices skyrocketing, this allows folks who can’t afford to fill all the memory slots on their motherboard to stick something in there that at least looks the part. This may seem pointless, but consider that many gamers and other power users have PC’s with clear side panels to show off their elaborate internal layouts. We get it from an aesthetic standpoint, but it also sounds like a new way to potentially get scammed when buying parts on the second-hand market. Though, to be fair, it could be that we’re just overly cynical after watching that Georges Méliès film. At the very least, the current price of memory certainly makes it feel like we’re being hit with a stick.
Finally, what good is living in a cyberpunk world without the occasional bout of rebellion? That’s where the Ageless Linux project comes in. This is a Linux distribution that’s intentionally configured to violate the California Digital Age Assurance Act, which essentially states that the operating system must ask the user how old they are and make this information available to any piece of software that wants to know.
To be fair, being in violation of this law right now is easy — indeed, the OS you’re using now is almost certainly not compliant. But the idea is that it may bend the knee at some point, while Ageless Linux won’t. One could argue that they started the project a bit too early, but frankly, the whole thing is performative in the first place, so if it gets people talking, that’s enough. We’re particularly interested in their idea of making a non-compliant hardware device that’s cheap enough to distribute while still meeting the definition of a computing device, as it’s written in the California Digital Age Assurance Act.
Think they would mind if we borrowed the idea for this year’s Supercon badge?
Advertisement
See something interesting that you think would be a good fit for our weekly Links column? Drop us a line, we’d love to hear about it.
If your Playdate wishlist is anything like mine (endless), here’s a good excuse to actually go ahead and free some of those games from limbo: Panic is running a sale across the Playdate Catalog to celebrate its three-year anniversary. Sure, the majority of Playdate games are pretty cheap as is, but they can still add up when you’re on a wild purchasing spree. Ask me how I know! The sale started on March 5 and goes until March 19 at 1PM ET (10AM PT), so take advantage of the discounts while you can.
There are 423 games available in the Catalog now, according to Panic, so if you’re having trouble deciding on which you should go for, I’ve got you covered with a few recommendations right here.
Season Two
If $39 felt like too much to drop on Season Two when it came out last summer, now’s the time to get it. Playdate’s second season had only half the number of games as its first, but it still felt like a much stronger collection. Each of its 12 games is really solid, and there’s plenty of variety in terms of genre and style, from puzzles and hours-long adventures to fast-paced action games that are great for bursts of intense play. And, it comes with Blippo+ — an oddball cable TV simulator that’s unlike anything out there right now.
All of these games are worth playing, but there were definitely some standouts from the bunch: The Whiteout, a post-apocalyptic adventure that’ll surely hit even harder now considering the winter we’ve had; the puzzle platformer Taria & Como; the arcade action game Fulcrum Defender; the climbing adventure, Tiny Turnip. I also really enjoyed Dig! Dig! Dino! for something on the chiller side.
Outside Parties
I have not been able to shut up about this game since it came out. It’s unique, it’s creepy, it’s completely engrossing and it really pushes the limits of what the Playdate can do. Outside Parties is a horror scavenger hunt, presenting you with one massive picture to scrutinize and find hidden scenes within, using the crank to adjust the brightness constantly so you can find anything that may be buried in shadow. As you find these targets, more and more of the game’s story comes to light through eerie audio clips. It is such a cool experience and the atmosphere of it all is incredible. You’ll get many hours of playtime out of this one too, with over 150 targets to find and lots of lore to uncover.
Crankstone
A full-blown western for the Playdate! Crankstone is a gallery shooter with minigames mixed throughout, and between the aesthetic, the music and the activities, it’s a lot of fun. You can choose the story mode to get right into the shooting and defending the town from outlaws, or head to the saloon to pick from the handful of mini games individually, including some fast-paced “spot the correct card” deck shuffling games and a few mimicry games involving the crank. It’s like a wild west theme park crammed into the Playdate, which is to say, it’s wonderful.
Echo: The Oracle’s Scroll
This is one of my all-time favorite Playdate games. Echo: The Oracle’s Scroll is a metroidvania without the usual combat, focusing entirely on exploration and puzzle-solving in a vast network of subterranean kingdoms. In this game, the Blight has forced civilization underground, and you play as a child who has been sent on a mission to deliver a scroll from the bottom-most territory, where the humans live, up to The Archives.
There are all sorts of treacherous environments underground, including magma lakes and areas filled with hostile vegetation, making for what is at times a challenging platformer that requires lots of creativity to make your way through. The tone is a bit somber, but quirky characters — like a frog prince with a bouncy belly — keep things from getting too dark.
Advertisement
Carte Blanche
This one’s for lovers of classic card games. There are six games in this virtual card game parlor (which is run by a bird named Blanche): Cribbage, Gin Rummy, Spades, Cassino and Spite & Malice. It’s great for if you already know what you’re doing, but I found Carte Blanche to be a really good introduction to these games for absolute beginners too, thanks to the easy-to-follow tutorials. When you win games, you’ll be rewarded with coins that you can spend at Blanche’s slot machine, which is stocked with little trinkets she’s collected.
Castle Kellmore
Castle Kellmore absolutely rules. This first-person action game puts you in a series of mazes where you have to fight off floating-head-style monsters as you hunt around for keys and try to find the doors and portals to your escape. There are sixteen levels, and upon finishing each one you’ll get a little summary of how long it took you to complete that area and what percentage of the level’s enemies you killed. I also really get a kick out of the sounds in this game. The enemies slurp and squelch, and your character will let out a hilariously passionless, “Ah” or “Ooeuugh” after picking up a health boost or getting injured. Great for fans of dungeons!
Piña Rollada
If you’ve ever played any of the Super Monkey Ball games, the gist of this one should be pretty familiar: roll the ball through the course and collect all of the fruits before reaching the exit. Don’t fall off the edge, and do it all in as little time as possible. Piña Rollada makes use of the Playdate’s accelerometer, which means you control the ball’s movement by tilting the console (there is also the option to use the D-pad instead). The courses start getting tricky pretty much right away, with thin paths that don’t have any guardrails, obstacles to avoid and moving platforms. And, just going near the exit will result in the ball getting sucked in, so you have to keep that in mind as you collect any surrounding fruits if you don’t want the level to end prematurely.
This is another one of those games that is both frustrating (in the fun way) and totally addicting. Expect to yell a lot.
Other games to try
These are just the games I’ve been enjoying lately, but there are tons of other Playdate games worth checking out during the sale, like these cheese games and Spilled Mushrooms. And if you need even more recommendations, take a look through our list of the best Playdate games, where you’ll find gems like Summit and Bwirds. There are quite a few I’m planning to finally spring from my wishlist too, including The Shape That Waits.
Update, March 15 2026, 7:15 PM ET: This story has been updated to include additional game recommendations.
Photo credit: Yanko Design Apple’s MacBook Neo has clearly struck a nerve, and it isn’t hard to see why people are already wondering what those same ideas might look like in a desktop form. Designer Sarang Sheth had exactly that thought, and the result is the Mac Neo concept, a compact desktop that takes everything that makes the laptop so compelling and reimagines it for the desk
This design is focused on sleekness, as it is noticeably thinner than Apple’s smallest existing desktop, allowing them to completely eliminate moving parts for cooling. Instead, they use basic passive airflow to keep it quiet, even if you’re hammering away at it for hours on end.
HELLO, MACBOOK NEO — Ready for whatever your day brings, MacBook Neo flies through everyday tasks and apps. Choose from four stunning colors in a…
THE MOST COLORFUL MACBOOK LINEUP EVER — Choose from Silver, Blush, Citrus, or Indigo — each with a color-coordinated keyboard to complete the…
POWER FOR EVERYDAY TASKS — Ready the moment you open it, MacBook Neo with the A18 Pro chip delivers the performance and AI capabilities you need to…
The color palette is quite welcoming and approachable, since they use the same four options as the laptop line, so you can have soft blush pink or citrus yellow, all in a smooth aluminum finish, which is a break from Apple’s customary plain metal tones. It gives the Mac Neo the impression that “this is a device that was picked, not just plopped on a desk,” and it would look just as good in your living room or dorm room as it would in the office.
The power comes from the same A18 Pro processor that powers the new laptop. So you’ll have six CPU cores, five GPU cores, and a dedicated neural engine to make it sing. Eigh gigabytes of unified memory makes everything zippy enough, and you won’t have to continuously swap files to and from storage. Testing the MacBook Neo indicates that it outperforms a slew of entry-level Windows machines in everyday use, and with so much more room to breathe in this desktop architecture, the processor stays nice and cool for longer.
They estimate the price to be roughly $400. This is primarily aimed for students, families, and anyone who is just getting into the Mac pool for the first time, putting it in line with many of the more basic laptops from other brands. As an aside, a separate power brick is included, which is beneficial because it allows them to manage heat better than if they put everything in a battery-powered laptop.
Looking for the most recent Strands answer? Click here for our daily Strands hints, as well as our daily answers and hints for The New York Times Mini Crossword, Wordle, Connections and Connections: Sports Edition puzzles.
Today’s NYT Strands puzzle is a fun and easy one if you love college basketball, and a certain big event that’s about to begin. Some of the answers are difficult to unscramble, so if you need hints and answers, read on.
If that doesn’t help you, here’s a clue: Tourney time
Clue words to unlock in-game hints
Your goal is to find hidden words that fit the puzzle’s theme. If you’re stuck, find any words you can. Every time you find three words of four letters or more, Strands will reveal one of the theme words. These are the words I used to get those hints but any words of four or more letters that you find will work:
MARK, BRACK, RACK, RACKS, CADS, CRAM, MOVE, MUTE
Answers for today’s Strands puzzle
These are the answers that tie into the theme. The goal of the puzzle is to find them all, including the spangram, a theme word that reaches from one side of the puzzle to the other. When you have all of them (I originally thought there were always eight but learned that the number can vary), every letter on the board will be used. Here are the nonspangram answers:
CHALK, BUBBLE, CINDERELLA, OVERTIME, BRACKET
Today’s Strands spangram
The completed NYT Strands puzzle for March 16, 2026.
NYT/Screenshot by CNET
Today’s Strands spangram is MARCHMADNESS. To find it, start with the M that’s four letters down on the far-left vertical row, and wind across.
To capitalize on Claude’s recent spike in popularity, Anthropic is offering a limited-time promotion that doubles usage limits for anyone using its AI chatbot during off-peak hours. From March 13 to March 27, users on Free, Pro, Max, and Team plans will get double the usage limits in a five-hour window when using Claude outside weekday hours between 8 AM and 2 PM ET. According to Anthropic, the promotion is automatic, and users don’t have to enable anything to get the benefits.
Anthropic said that this promotion applies to anyone using Claude on web, desktop or mobile, but also with Cowork, Claude Code, Claude for Excel and Claude for PowerPoint. Previously, Anthropic offered a similar event from December 25 to December 31, doubling usage limits for Pro, Max 5x or Max 20x subscribers. However, Anthropic is targeting an even wider audience with its latest promotion since only Enterprise users are excluded this time around.
Anthropic is marketing the promotion as a “small thank you to everyone using Claude,” but it’s likely tied to its ongoing battle with the Department of Defense. After refusing to remove certain AI safeguards for the Department of Defense, Anthropic was listed as a supply chain risk and lost its contract with the federal agency. In turn, OpenAI signed a deal with the Department of Defense, leading to many users deciding to boycott ChatGPT in favor of Claude and other AI chatbot options.
The eagle-eyed will have spotted I predicted a Sinner Best Picture win and Paul Thomas Anderson Best Director win. Am I predicting that rare beast of a split for the top prizes or cowardly splitting my chances? Both.
But, let’s not forget Sinners leads with an incredible 16 nominations. Sixteen!
Advertisement
To put that in perspective that’s the most in Oscar history by not one but two nominations, albeit this includes a new category Best Casting but it’s incredible. A film which came out nowhere near the typical awards season window towards the end of the year and about bloody vampires is the FRONTRUNNER for the Academy Awards.
However, despite that and maybe due to the unstoppable juggernaut which One Battle After Another has been most of awards season with 13 nominations, it’s also managed to position itself as the scrappy underdog.
It would be a sin for any other film to win, right?
Advertisement
Cards on the table
For full transparency, this amateur film writer has attempted to predict the categories in what is called the most unpredictable Oscars in years (they always say that though).
Here’s my predictions so you can hold me to account later…
Best Picture – Sinners
Best Director – Paul Thomas Anderson
Advertisement
Best Actor – Timothée Chalamet
Best Actress – Jessie Buckley
Best Supporting Actor – Sean Penn
Best Supporting Actress – Teyana Taylor
Advertisement
Best Original Screenplay – Sinners
Best Adapted Screenplay – One Battle After Another
Best Animated Feature – KPop Demon Hunters
Best International Feature – Sentimental Value
Advertisement
Best Documentary Feature – The Perfect Neighbour
Best Documentary Short – All the Empty Rooms
Best Live Action Short – Two People Exchanging Saliva
Best Animated Short – Butterfly
Advertisement
Best Original Score – Ludwig Göransson
Best Original Song – Golden
Best Sound – F1
Best Casting – Sinners
Advertisement
Best Production Design – Frankenstein
Best Cinematography – Sinners
Best Makeup and Hairstyling – Frankenstein
Best Costume Design – Frankenstein
Advertisement
Best Editing – One Battle After Another
Best Visual Effects – Avatar: Fire and Ash
One Sinner After Another
Not since the year of Barbenheimer have two films completely dominated the Oscar conversation. One Battle After A Sinner? One Sinner After Another? Hmmmm there’s not really a good portmanteau for them…
Advertisement
It’s pretty much the conventional wisdom that either Sinners or One Battle After Another will bring home the big prizes of Picture, Director, Writing and some of the other acting categories.
But who will prevail?
How to watch the Oscars wherever you are…
Whether you’re strapping in as we approach midnight with a strong coffee in Europe…or you’re watching as sunset rises with a strong coffee in Asia (that’s where this writer is) or you’re watching at a sensible time with a strong coffee in America…how do you watch the ceremony?
Advertisement
Well, thankfully we’ve got you covered! Find out how to watch it wherever you are with our handy guide here.
What does the ceremony look like?
For those new to Hollywood’s sacred annual ritual, here’s what the night looks like…
24 categories will be presented include this year’s newest one, Best Casting, just to boost that running time a little more.
Advertisement
You have the much covered acting categories – with Actor, Supporting Actor, Actress and Best Actress – the more technical awards – Production Design, Sound etc – and then the big ones: Writing, Directing and Picture.
Who will take home the most Academy Awards, stay tuned…
Welcome to the Oscars!
Hello from wherever you are in the world, and welcome to the biggest night in Hollywood!
Advertisement
Strap in for what’ll most likely be – if we’re being conservative judging by the ‘shortest’ ceremonies – one hour of pre-coverage and four hours of Oscar drama. And I’ll be with you through it all.
Stay tuned wherever you are as the action kicks off in less than an hour…
Please follow our community guidelines.” data-join-the-conversation-text=”Join the Conversation”>
You must confirm your public display name before commenting
Advertisement
Please logout and then login again, you will then be prompted to enter your display name.
Customers at Apple Grand Central are being directed to other stores while it’s closed for March 12 and 13. There’s no explanation, but it’s likely to be to do with Apple’s 50th anniversary celebrations.
View of Apple Grand Central from across the station — image credit: Apple
Apple Grand Central is used to being used for promotional events — it was where the “Severance” pop-up was in January 2025. Given its size and how many people go by it in Grand Central Station, it would make sense for an anniversary event to be held there. As yet there is no indication, though, of whether it’s an anniversary event, an unrelated promotion, or simply a refurbishment of the store. Buyers coming to the store on Thursday March 12, 2026, just saw a sign telling them it was closed. Continue Reading on AppleInsider | Discuss on our Forums
In an unexpected twist, humans have taken some jobs back from AI. Embark Studios’ CEO Patrick Söderlund recently told GamesIndustry.biz that the studio “re-recorded” some of the AI-generated voice lines in Arc Raiders with human voices, only after its successful launch in October.
“There is a quality difference,” Söderlund told GamesIndustry.biz. “A real professional actor is better than AI; that’s just how it is.”
With Arc Raiders’ player count peaking at nearly half a million users on Steam, the game’s breakout success was still marred by its use of text-to-speech AI. While there was no generative AI used for the visuals of the extraction shooter, Embark Studios paid its actors for approval to license their voices for text-to-speech AI, according to Söderlund. Even though Söderlund said that the text-to-speech AI was reserved for lines “that aren’t as essential to the immersion of the experience,” many players weren’t happy with this creative decision.
Responding to the criticism, Embark Studios is seemingly reversing course and relying more on its voice actors. Söderlund said that the studio pays its voice actors for their time in the recording booth and will “continue to bring many of them back as we carry on updating the game.” However, it’s important to note that Söderlund told GamesIndustry.biz that “some” of the AI-generated lines were replaced by voice actors, which could indicate that the studio isn’t looking to completely ditch its text-to-speech AI anytime soon.
Local AI inference gains attention as cloud costs continue rising
Developers increasingly explore running language models directly on personal hardware
External GPU enclosures have existed for some time – typically associated with gaming laptops and graphics acceleration tasks that exceed the capabilities of mobile processors.
Plugable’s newly released TBT5-AI belongs to this category, but introduces a design focused on connecting desktop graphics hardware to laptops for local AI workloads.
The enclosure provides a full-length PCIe x16 slot that allows users to install a desktop-class graphics card inside the external chassis.
Article continues below
Advertisement
Desktop-class hardware in an external enclosure
An integrated 850-watt power supply delivers the energy required to run high-performance GPUs that would normally only operate inside desktop workstations.
For connectivity, this device comes with a single Thunderbolt 5 cable, which permits direct connection with a laptop, and supports up to 80Gbps of bidirectional bandwidth, while a boost mode can increase throughput to 120Gbps for certain workloads.
Advertisement
Inside the enclosure, this bandwidth links the installed GPU through PCIe 4.0 x4 lanes, reducing the transfer bottlenecks that limited earlier external GPU designs.
In addition to housing the graphics card, the system functions as a hub that expands connectivity for the attached laptop.
Sign up to the TechRadar Pro newsletter to get all the top news, opinion, features and guidance your business needs to succeed!
It delivers up to 96 watts of charging power while also providing 2.5-gigabit Ethernet networking and several high-speed USB ports.
Advertisement
According to Plugable, many engineers increasingly want to keep model processing and data handling within their own systems, and the TBT5-AI offers just that, as it is designed for developers experimenting with local AI inference environments.
The device allows developers to run large language models directly on local hardware instead of sending workloads to cloud infrastructure.
It supports common local AI frameworks, including llama.cpp, Hugging Face models, and Nvidia’s NIM inference platform.
Advertisement
Plugable chief technology officer Bernie Thompson said the hardware targets industries where protecting sensitive information remains a strict operational requirement.
“Data privacy is not a feature but a mandate,” Thompson said, referring to sectors such as healthcare, financial services, and legal organizations.
Plugable is also preparing enterprise versions dubbed TBT5-AI16, TBT5-AI32, and TBT5-AI96 that will include bundled graphics processors.
These configurations will integrate a software environment called Plugable Chat, described as an air-gapped AI orchestration platform for regulated organizations.
Advertisement
The company claims that these systems will shift AI processing away from subscription-based cloud services toward locally controlled computing infrastructure.
Priced at $599.95 as a standalone unit, the Plugable TBT5-AI enclosure was officially released a few days ago, and it is now available via Amazon and Plugable.com.
