A year ago the Ryzen AI Halo, AMD’s tiny new AI workstation, would have offered devs and machine learning enthusiasts an Nvidia DGX Spark-like experience at a fraction of the cost.
Unfortunately for AMD, time and the ongoing memory shortage, which both AMD itself and Nvidia are partially responsible for, hasn’t been kind to the consumer electronics industry.
Launching at a hair under $4,000, the AI Halo is still cheaper than the Spark at its new MSRP of $4,699, but is now a much tougher sell than when you could get the same hardware for as little as $2,000.
That’s right. The 128 GB AI Halo is based on year-old technology. Its main selling point, and what AMD has spent the past several months getting right, is the packaging. Much like with the Spark, you’re not just buying the machine but all the software and documentation you need to run and fine-tune enterprise-grade models and AI agents like OpenClaw and Cline, locally.
Many will, understandably, balk at the price — $4,000 is a down payment on a car — the system is still one of the most affordable options for those who need more than the 32 GB that the highest-end graphics card can provide.
Not long ago, building a workstation with 128 GB of video memory would have set you back at least $20,000, and that was before the RAMpocalypse. This puts systems like DGX Spark and AI Halo in a rather unique position.
The Ryzen AI Halo was clearly inspired by the DGX Spark. Measuring in at 5.9 x 5.9 x 1.79 inches, the black and silver system shares a nearly identical form factor to its competitor.
Rather than gold aluminum siding, AMD has opted for a more subdued look with a textured top cover adorned by its logo and an LED light bar that wraps around its perimeter. The chassis itself is well ventilated with intake located along the front of the system sides and heat exhausting out the back.
The back of the system is adorned with four USB-C ports, one of which is dedicated to power, while the remaining three offer connectivity (1x USB 3.2, 2x USB 4.0) for storage and peripherals. The AI Halo supports display out on all three of those ports as well as via HDMI 2.1b . A single RJ45 network port provides 10 Gbps of connectivity for those who prefer wired connectivity over the onboard WiFi 7 radio.
One thing you won’t find on the back of the AI Halo are QSFP ports for high-speed networking. The DGX Spark features a 200 Gbps ConnectX-7 SmartNIC for clustering multiple devices together. The AI Halo does still support clustering if you happen to pick up multiple systems, but with only one such system on hand, we can’t say how big a difference the slower networking actually makes.
AMD’s Ryzen AI 395+, which you may recognize from its codename Strix Halo, sits at the heart of the system. This SoC isn’t new, having been on the market for more than a year now. In fact, we pitted the Pro variant of the chip running in HP’s Z2 Mini against the DGX Spark’s GB10 SoC back in December 2025.
The chip is equipped with 16 Zen 5 cores clocking up to 5.2 GHz along with an RDNA 3.5 GPU with 40 compute units putting out around 56 teraflops of dense FP16 performance under ideal conditions.
While Strix Halo can be obtained with as little as 32 GB of LPDDR5X memory, the AI Halo is packing 128 GB as standard. That’s enough to run models of up to 200 billion parameters in size, at 4-bit precision that is. Out of the box, our system was configured to share up to 75 percent, or about 96 GB, of that with GPU. However, at least on Linux, you can extend this to nearly the system’s full capacity.
That memory is connected to the SoC by a 256-bit bus good for about 256 GB/s of bandwidth — more than you’d get on a (non-Pro) Threadripper system.
Bandwidth is a major bottleneck for LLM inference, with token generation directly proportional to how fast the memory actually is, and because the AI Halo’s memory hangs off the GPU, it can take full advantage of it.
While 256 GB/s is a lot for DDR5, it is dwarfed when you compare to the GDDR or HBM found in consumer and datacenter GPUs. The RTX 5090 delivers 1.7 TB/s of bandwidth, making it admittedly high — for models small enough to fit in that card’s 32 GB of VRAM.
We’ll talk about performance in a bit, but this really gets to the hardware’s core value proposition. For most local AI enthusiasts and devs, memory capacity is the biggest bottleneck.
It doesn’t matter how many teraflops your GPU can push or how fast your memory is, if you don’t have enough of it in the first place. At 16-bit precision you need about 2 GB of memory for every billion model parameters. At 8-bits, it’s a 1:1 ratio and, at 4-bits, you need just 512 MB for every billion parameters.
If you’ve toyed around with local LLMs in Ollama or LM Studio before you’re almost certainly running 4-bit weights, which is why you can cram a 20 billion parameter model onto a consumer graphics card with as little as 16 GB of VRAM.
Unfortunately, there are a lot of AI workloads that aren’t easily quantized or require substantial quantities of memory in addition to what’s used to hold the model weights. But once you venture beyond low precision inference, memory quickly becomes a major constraint. For example, a full fine tune of a modest 7B parameter can easily consume upwards of 100 GB of memory.
This is where systems like the AI Halo or DGX Spark really shine. They may not be the most powerful or the fastest systems, but there’s not much that you’d want to do that you couldn’t thanks to their ample memory capacity.
As we’ve shown in the past, Strix Halo is more than capable of running larger more capable models exceeding 100 billion parameters or fine-tuning models up to 70 billion parameters, something that’s well beyond the means of consumer graphics cards.
If the chip isn’t new, you might be wondering what exactly the Ryzen AI Halo buys you over another Strix box, like HP’s Z2 Mini G1a we reviewed back in December. Back then, that system retailed for around $3,000. Its price has since surged to nearly $4,900.
If you’re already familiar with AMD’s HIP and ROCm stacks and reasonably comfortable with Linux, the answer is not a lot. AMD even has playbooks specifically for early adopters of its Ryzen AI products. So, if you jumped on a Strix Halo system before DRAM prices hockey sticked, you’re really only missing out on the conveniences that the preinstalled software brings.
With that said, we’re willing to bet most folks considering AI Halo are probably dipping their toes into ML and AMD’s software ecosystem for the first time.
ROCm is a heck of a lot easier to get running on Ryzen APUs and Radeon graphics than it used to be, but we’d be lying if we said that it’s always easy. The same is true of Nvidia and CUDA to a lesser extent. Some steps are easier, while others like GPU passthrough for containers require jumping through additional hoops.
That’s not even to mention PyTorch compatibility, which can vary from app to app. Regardless of which platform you buy into, wrangling dependencies is still a mess.
Both the AI Halo and DGX Spark’s core value prop is helping customers avoid as many of these headaches as possible by combining validated hardware with pre-installed dependencies and well documented playbooks that walk you through common use cases.
In other words, it’s an AI lab in a box.
The AI Halo ships with your choice of Linux or Windows 11. The review unit AMD provided us with, came equipped with a lightly-modified version of Debian with the 6.18 Linux Kernel, Gnome desktop environment, ROCm 7.13 preinstalled, and a slew of preinstalled AI apps and frameworks, like ComfyUI and vLLM.
For anyone who’s used Linux before, the experience should be quite intuitive. Upon first boot, a startup wizard will guide you through the process of creating your user profile, connecting to the network, and updating the device.
Once you are logged in, AMD’s Ryzen AI Developer Center launches automatically and provides quick access to resources and system settings.
As of this writing, AMD’s developer docs include 19 playbooks covering everything from the basics of running LLMs and image models on the device to building full blown agents with OpenClaw.
We walked through most of these as part of our review process and with a few exceptions we were able to run them with minimal troubleshooting. We did have to ask an LLM for help debugging AMD’s PyTorch fine-tuning scripts. Thankfully, the selection of pre-downloaded models were capable enough to identify the single line fix required to get them running again.
While most of AMD’s playbooks were more than adequate, we found its vLLM getting started guide a little lacking. It was easy enough to get it running —AMD has written a wrapper that abstracts the creation and deployment of the inference server in a Docker container — but the guide doesn’t discuss how to select a model, much less configure it.
vLLM is an incredibly popular inference server broadly deployed in production. This makes it all the more disappointing that AMD’s documentation isn’t more comprehensive.
One bright spot we’d like to highlight is Lemonade Server. The app comes preinstalled and provides an LM Studio or Ollama-like experience tuned specifically for AMD hardware. It’s built atop a number of different model runners including vLLM, Llama.cpp, Whisper.cpp, Stable Diffusion.cpp and others. There is even support for a limited selection of models which will run on the system’s NPU.
Perhaps the most attractive use case for the system is as a host for AI agents.
When AMD announced the system, it was keen to highlight how small local models, like Qwen 3.6-35B-A3B, were now good enough to replace larger proprietary models for many coding workflows.
The company went so far as to claim that, for full-time software devs, the system could save as much as $750/month compared in API expenses they’d pay to a cloud-based LLM. We plan to put those claims to the test in a future article. Beyond AI coding, we also expect the system to be quite popular as a platform for running harnesses like OpenClaw.
Given the software’s significant, not to mention numerous security implications, running it locally with container isolation is probably the safest option, and its large memory capacity means that you’ll have access to larger more capable models.
In terms of performance, the Ryzen AI Halo is a bit of a mixed bag.
In memory bound applications like LLM inference, the system matches and in some cases narrowly outpaces Nvidia’s more expensive DGX Spark. Hanging the memory off the GPU instead of the CPU benefits the AI Halo here.
In compute bound workloads, like fine tuning, image generation, or batch processing, the gap grows considerably. We plan to dig deeper into how the AI Halo performs in a future article, but, in our initial testing, we don’t see a major uplift in performance compared to our earlier testing.
We’re also not sharing vLLM performance figures for the AI Halo just yet as our initial testing with AMD’s provided build produced results we’re not confident in.
Depending on the workload and precision, you can expect the Spark’s GB10 APU to be anywhere from 2x to 3x quicker in compute-bound AI workloads.
A big piece of this is down to the fact that Strix Halo wasn’t really intended for this use case. AMD’s RDNA 3.5 GPU tech lacks support for floating point precisions lower than FP/BF16. It does offer INT8 support, but only by upcasting to FP16, which means no performance uplift from dropping to lower precision.
On paper, the GB10 delivers roughly twice the 16-bit performance, three times that at FP8 and twice again at FP4. This is one of the biggest critiques of AMD’s current consumer hardware roadmap, and why we continue to see such a wide performance delta. While its software has improved and its datacenter kit supports FP8 and FP4, the AI Halo is stuck on an older microarchitecture.
But, as we mentioned in our initial Strix Halo vs GB10 head-to-head, whether you’ll even notice the performance deficit really depends on what you’re doing.
AI benchmarks, including ours, usually disable prefix caching. This allows us to accurately evaluate the accelerators’ performance, but isn’t representative of how you’d actually use the model.
In a chatbot or AI agent, the prefix caching keeps the accelerator from getting bogged down by caching previously-computed information so that only new data has to be processed. With it disabled, the problem size grows with each prompt processed and each response generated.
We’re currently in the process of developing a series of new tests that take advantage of caching and other functionality, like multi-token-prediction to measure performance in agentic applications like code generation. We look forward to sharing the results of those tests soon.
Strix Halo wasn’t a cheap part before RAMageddon and it certainly isn’t now — $4,000 is a lot of money. But for the right person, it’s still a relative bargain.
If you’re interested in learning more about local AI, we recommend starting with what you’ve already got before considering dropping this kind of cash on an AI-first system like the AI Halo.
If usage based APIs are out of the question and your existing graphics card is no longer cutting it, GB for GB the Ryzen AI Halo is still much cheaper than workstation cards from either AMD or Nvidia. For reference, a 96 GB RTX Pro 6000 is much, much faster and offers nearly as much addressable memory as the AI Halo, but has an MSRP of $13,250. Oh, and that’s just for the GPU; you still need to plug it into something with at least that much DDR5 on board.
And so the question becomes how badly do you need the VRAM and how valuable is AMD’s documentation and support? Enthusiasts willing to blaze their own trail might be able to save a buck by picking up an OEM Strix Halo box and configuring it themselves.
On the flip side, for those willing to spend a bit more money, Nvidia’s DGX Spark also offers fantastic documentation and a fair bit more computational grunt, which again means faster fine tuning, image generation, and prompt processing. The number of tok/s is limited by memory bandwidth.
With that said, the DGX Spark is much more of an appliance, which means, if you buy this thinking you’re going to run agents on it and later decide it’s not worth the trouble, it’s less likely to end up collecting dust on the shelf. Because it’s just an x86 PC under the hood, the AI Halo is perfectly capable of running Windows or your preferred Linux distro, if you decide local AI just isn’t for you.
Oh, and if 128 GB of VRAM isn’t enough for you, AMD has a refreshed version of the system on the way with 192 GB of LPDDR5X memory and slightly higher clocks. ®
Last week, researchers at cloud security firm Sysdig said they’d documented the first known case of “agentic ransomware.” It was an extortion operation, dubbed JadePuffer, in which an AI agent — not a human — handled the technical execution of a real-world cyberattack from start to finish. The agent broke into a vulnerable server, stole credentials, moved through the target’s network, encrypted files, and even wrote its own ransom note, adapting to obstacles along the way like a human hacker would. Coverage of the funding described it as run “without any human oversight,” with “no human at the keyboard.”
That’s not quite the full picture. In an interview on Monday with CyberScoop, Sysdig’s Michael Clark, the company’s senior director of threat research, clarified that a human was still very much involved — just not in the technical execution. “A human still set up and pointed the operation and provisioned the infrastructure behind it, the command-and-control server, the staging server used for the stolen data and chose a victim,” Clark said. The credentials used to break into the victim’s database, he added, weren’t harvested by the AI agent itself; someone obtained them separately, through a prior compromise, and handed them to the operation.
None of this contradicts Sysdig’s original claim, and the technical details of the attack remain notable on their own — wild, even. The agent got in through a known bug in Langflow, a popular open-source tool for building LLM apps, then moved on to a production MySQL server and exploited another known flaw to gain admin access. It encrypted over 1,300 configuration records and not only left behind a ransom note that it wrote itself but it left a Bitcoin address where the ransom could be sent. Sysdig hasn’t disclosed who was targeted.
The techniques were fairly ordinary apparently, what stood out was the speed and transparency involved. The agent fixed a failed login in 31 seconds, narrating its own reasoning in natural-language code comments the whole way.
One detail that initially seemed to muddy the picture has since been clarified. Clark had told CyberScoop that Sysdig found “multiple models were used in the attack,” citing harvested keys for OpenAI, Anthropic, DeepSeek, and Gemini — language that left open the question of whether several models actively powered different stages of the intrusion. Asked to clarify, Clark told TechCrunch that those keys were simply part of what the agent stole, not evidence of what was driving it.
“The agent swept the Langflow host for anything valuable — provider API keys, cloud credentials, cryptocurrency wallets, and database configs — and those provider keys were part of the loot,” he said via email. “They are indicative of what the attacker considered worth taking, but they do not tell us which model was making the decisions.”
On the model actually running JadePuffer, Clark said Sysdig “was not able to identify the specific model driving the agent” and has no visibility into its system prompt or configuration.
Microsoft researcher Geoff McDonald’s theory, offered on LinkedIn several days ago, is worth revisiting in that light. McDonald suspected an open-weight model with safety training stripped out, rather than a frontier model, was behind the attack, based on his own red-teaming experience showing frontier labs’ safety layers hold up well. Sysdig’s own account doesn’t confirm or rule that out.
McDonald’s post also warned that ransomware campaigns are now bounded primarily by attacker budget rather than human effort, raising the possibility of “thousands or tens of thousands of simultaneous campaigns.” That concern is a little harder to square with what Clark described Monday. (If a human still has to choose each victim, provision infrastructure, and obtain database credentials for every operation, that’s a bit of a bottleneck, at least.)
Either way, Clark told CyberScoop, while Sysdig hasn’t seen the same operation hit other victims yet, given how cheap it is to run an agent, he expects that to change.
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.
Security firm Sysdig says it documented the first fully agentic ransomware attack, dubbed JadePuffer, in which an AI agent planned and executed an entire database-extortion operation with no human at the keyboard. The agent exploited a Langflow flaw, moved laterally, encrypted 1,342 config items, and diagnosed a failed login in 31 seconds, but never saved the decryption key, making recovery impossible.
TL;DR
Security firm Sysdig says it has documented the first ransomware attack run end to end by an AI agent, first reported by Business Insider. A large language model planned, executed, and adapted the entire operation, which Sysdig has named JadePuffer.
The agent chained together every stage of the attack, from reconnaissance and credential theft to lateral movement and data encryption. It did so with no human directing the keyboard, according to the company’s threat research team.
The intrusion began by exploiting a known Langflow remote-code-execution flaw to harvest cloud and AI-provider credentials. The agent then compromised a production database, encrypting 1,342 configuration items and leaving a ransom note.
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!
The clearest sign of autonomy came when an admin login failed, and the agent diagnosed the problem and issued a working fix in 31 seconds. More than 600 payloads across the campaign carried plain-language comments explaining the agent’s own reasoning, per BleepingComputer.
There is a grim catch for any victim. The ransom note’s decryption key was reportedly never saved, making recovery impossible even if the ransom were paid.
The significance is less the sophistication than the deskilling, as an agent can now chain steps that once demanded expertise at each one. Ransomware is edging from a craft into a prompt.
JadePuffer does not stand alone. Anthropic recently disrupted what it called a largely AI-run cyber espionage campaign, and Google identified the first AI-developed zero-day exploit before it could be used at scale.
Governments are alarmed, with the UK’s Yvette Cooper warning of an AI “Hiroshima” without rules and frontier labs racing China on offensive capability. The same models that can be coaxed into misbehaviour are now cheap enough to weaponise.
Defenders are not standing still, and the industry is betting that 2026 becomes the year of governed cybersecurity AI. The uncomfortable truth is that both sides now field the same tools.
Sysdig’s case is a proof of concept as much as an incident, since one working example tends to become a template. The keyboard is empty, but the attack still runs.
Amazon has cut a total of 57 jobs in Washington state across various teams, including roles at the director and senior manager levels, according to a filing made public Monday morning.
People impacted by the cuts include 16 software engineers as well as product managers and creative marketing employees working in Seattle and Bellevue offices. Nine remote employees, including investigation specialists and risk managers, were also let go.
Employees were notified of the layoffs throughout May and in early June, according to an Amazon filing with the Employment Security Department, released Monday under the Worker Adjustment and Retraining Notification (WARN) Act. The roles are scheduled to end in August.
“[W]e filed a WARN notice because a few businesses across the company made organizational changes that each impacted a small number of employees — in most cases fewer than five employees per business,” said Brad Glasser, an Amazon spokesperson, via email.
WARN notifications are triggered by state law when more than 50 Washington-based employees in total are laid off over a period of 30 days.
“We don’t make decisions like this lightly, and we’re committed to supporting the employees who were impacted,” Glasser added.
It’s a sign of the broader belt-tightening across the tech industry. Microsoft separately cut more than 600 jobs in Washington state on Monday morning, part of global layoffs eliminating 4,800 roles across the Redmond company, primarily in sales, consulting and gaming.
The latest Amazon cuts follow layoffs of 2,198 Washington-based employees in February and 2,303 in October 2025. Globally, the company has eliminated roughly 30,000 positions in the past year, cumulatively amounting to the the largest workforce reduction in its history.
The multiple rounds of layoffs have hit wide-ranging positions and divisions, with software engineers the hardest hit. Corporate support, commercial functions, legal, tax, and ad sales positions have all seen cuts, as have Amazon’s core technology organization, gaming division and robotics unit.
The previous larger cuts were part of an effort to “reduce layers, increase ownership, and remove bureaucracy,” according to a memo sent to employees and posted online earlier this year by Beth Galetti, senior vice president of people experience and technology.
Amazon’s corporate roles numbered around 50,000 in the Seattle area.
Tech giants nationwide have made round after round of job cuts in the past year as they pour billions into AI data center expansions and gain labor efficiencies through the use of artificial intelligence.
Amazon reported $181.5 billion in sales for the first quarter of this year, up 17% from a year earlier. Profits came in at $30.3 billion, boosted by gains tied to the value of its investment in Anthropic.
Image generation features found in Apple Creator Studio rely on Google Cloud servers, but users will be warned before prompts are sent to the third-party AI tool.
Apple Intelligence is powered by Apple Foundation Models found on your iPhone and in Apple’s Private Cloud Compute servers. Those are distinct features and models from the integrations that utilize third-party AI tools like Google Cloud and ChatGPT.
After updating to the latest Apple Creator Studio version, users are encountering a new pop-up, whether they are running iOS 26 or iOS 27. That pop-up warns that the user’s prompt will be sent to a Google Cloud server, but won’t be used for training.
The warning is similar to what would appear when user queries were being sent to ChatGPT in previous versions of Apple Intelligence.
To be perfectly clear: This is not a part of Apple Intelligence or Apple Foundation Models.
There has already been some confusion around this new warning and Apple’s work with Google to implement Gemini technology in the new Apple Foundation Models. The Apple Foundation Models and resulting Apple Intelligence and Siri AI upgrades do not use any Google services, Google Search, Gemini Assistant, or Google frameworks.
The Apple Foundation Models on your device and in Apple’s Private Cloud Compute servers are Apple technology all the way down. Yes, the new models were built with Gemini Frontier models and servers at the foundation, but nothing Google remains in the shipping models.
Apple is working to bring its most powerful Apple Foundation Models to Google servers with Nvidia GPUs, but via Private Cloud Compute. Those Google servers Apple uses for Private Cloud Compute are fully Apple’s in operation, just like iCloud servers are when using AWS.
When you go to generate a shape or image in Pages, Freeform, or any other Apple Creator Studio app that has these features, it is using Google Cloud. Users have the ability to accept the warning each time, or set it to always accept.
The only data being sent in these instances is the text you’ve typed in the prompt or image sent to edit. And even then, just like with OpenAI’s partnership, Google is unable to train on sent prompts or retain data from the interaction.
Third-party AI usage limits in Apple Creator Studio
The feature is wholly isolated to Apple Creator Studio, so if a user would prefer to avoid using Google Cloud, it is easy to do so. Although, those that do choose to use it can know that their data remains private for the interaction.
Since Apple Creator Studio AI features rely on external AI tools, there are limitations to what can be done. Apple shares the percentage of AI usage in app settings, and that usage gets reset each month.
OpenAI provides ChatGPT for slide generation, and users can generate about 50 presentations with 8-10 slides each with their allotment. Google Cloud can generate 50 images or 250 shapes with its monthly allotment.
Apple doesn’t specify how many tokens a user has nor how many an event expends. It’s up to the user to keep queries short to minimize use, and to monitor usage manually.
The support document defining these features reiterates that zero data is used for training models.
Flight sims are wonderful to play around with to get immersed in the position of a pilot. Racing sims can give you a thrill that can only be beaten by the real thing. However, most of this tech is on the more expensive side, so it would be great if you could use some of the hardware already found in your house. Many Sony headphones already have rotation and movement data built in for spatial audio, so why not start there?
[Nicholas Slattery] had this very idea and has produced an open-source application to connect your headphones straight to your sim. There’s a surprising amount of support built into many headsets that use a known protocol called the Android Head Tracker HID protocol. This allowed [Nicholas] to connect a family of Sony headphones straight into OpenTrack, which is often used with flight sims. The best part is you can still use the headphones as normal with a Bluetooth connection.
If you want to give this a try with your own rig, check out [Nicholas]’s GitHub here. While flight and driving sims might be expensive to put together, it’s never too hard to hack together something to lower that barrier! Whether it’s a flight sim force-feedback joystick or driving sim hand-breaks we got you!
Looking for a different day?
A new NYT Connections puzzle appears at midnight each day for your time zone – which means that some people are always playing ‘today’s game’ while others are playing ‘yesterday’s’. If you’re looking for Monday’s puzzle instead then click here: NYT Connections hints and answers for Monday, July 6 (game #1121).
Good morning! Let’s play Connections, the NYT’s clever word game that challenges you to group answers in various categories. It can be tough, so read on if you need Connections hints.
What should you do once you’ve finished? Why, play some more word games of course. I’ve also got daily Strands hints and answers and Quordle hints and answers articles if you need help for those too, while Marc’s Wordle today page covers the original viral word game.
SPOILER WARNING: Information about NYT Connections today is below, so don’t read on if you don’t want to know the answers.
Today’s NYT Connections words are…
What are some clues for today’s NYT Connections groups?
Need more clues?
We’re firmly in spoiler territory now, but read on if you want to know what the four theme answers are for today’s NYT Connections puzzles…
What are the answers for today’s NYT Connections groups?
Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON’T WANT TO SEE THEM.
The answers to today’s Connections, game #1122, are…
The game of Clue has come up a couple of times in Connections — previously we had weapons and characters — and although I’ve never played it I was familiar with Colonel Mustard and somehow knew that a candlestick could be used as murder weapon.
This time, though, I got ROOMS IN CLUE just because they were also, erm rooms.
My mistake came in trying to make sense of the final eight tiles, first making an utterly random quartet and then taking some time and spotting Bert, Ernie, Count and Elmo.
NYT Connections is one of several increasingly popular word games made by the New York Times. It challenges you to find groups of four items that share something in common, and each group has a different difficulty level: green is easy, yellow a little harder, blue often quite tough and purple usually very difficult.
On the plus side, you don’t technically need to solve the final one, as you’ll be able to answer that one by a process of elimination. What’s more, you can make up to four mistakes, which gives you a little bit of breathing room.
It’s a little more involved than something like Wordle, however, and there are plenty of opportunities for the game to trip you up with tricks. For instance, watch out for homophones and other word games that could disguise the answers.
It’s playable for free via the NYT Games site on desktop or mobile.
Looking for a different day?
A new NYT Strands puzzle appears at midnight each day for your time zone – which means that some people are always playing ‘today’s game’ while others are playing ‘yesterday’s’. If you’re looking for Monday’s puzzle instead then click here: NYT Strands hints and answers for Monday, July 6 (game #855).
Strands is the NYT’s latest word game after the likes of Wordle, Spelling Bee and Connections – and it’s great fun. It can be difficult, though, so read on for my Strands hints.
Want more word-based fun? Then check out my NYT Connections today and Quordle today pages for hints and answers for those games, and Marc’s Wordle today page for the original viral word game.
SPOILER WARNING: Information about NYT Strands today is below, so don’t read on if you don’t want to know the answers.
• Today’s NYT Strands theme is… Hitching a ride
Play any of these words to unlock the in-game hints system.
• Spangram has 10 letters
• First side: left, 5th row
• Last side: right, 5th row
Right, the answers are below, so DO NOT SCROLL ANY FURTHER IF YOU DON’T WANT TO SEE THEM.
The answers to today’s Strands, game #856, are…
I was having a merry old ride around this board until the final word and spangram slowed me down to a standstill.
Spangrams like IGETAROUND really are my nemesis and I came close, but not that close, to using a hint, but thankfully saw the unseasonal SLEIGH just in time before I disappeared into a letter-soup haze.
Finding non-game words was actually a good deal harder and I struggled to find the six required for this page.
Strands is the NYT’s not-so-new-any-more word game, following Wordle and Connections. It’s now a fully fledged member of the NYT’s games stable that has been running for a year and which can be played on the NYT Games site on desktop or mobile.
I’ve got a full guide to how to play NYT Strands, complete with tips for solving it, so check that out if you’re struggling to beat it each day.
With the latest iOS 27 developer beta, Apple is giving testers an early look at one of the upcoming improvements to its AI-powered Siri: the ability to adjust how quickly and expressively the AI assistant speaks. In iOS 27 beta 3, out today, Apple has enabled the voice controls for “Pace” and “Expressivity” that were previously labeled as “Coming soon” in the first developer beta releases.
The update is part of Apple’s broader effort to make Siri feel more natural and personal, as it rebuilds the assistant around generative AI. Like ChatGPT and others offering voice AI assistants, letting users customize how the AI sounds is an important aspect in helping connect people with the new technology.
However, ChatGPT’s voice-customization options allow users to go even further, as the ability to adjust the AI’s warmth and enthusiasm was rolled out in December 2025, alongside options to configure the base style and tone. The latter lets users adjust OpenAI’s assistant to be more friendly, professional, candid, or quirky, among other styles. This is reflected not only in how ChatGPT speaks, but also in how it presents information to the user.
First introduced at Apple’s Worldwide Developers Conference (WWDC 26) in June, Siri’s voice controls let users personalize their Siri experience beyond just choosing a male- or female-sounding assistant. Now beta testers will be able to switch between a range of voices with different accents, and then use sliders to change how slowly or quickly Siri speaks and how much human-like emotion its voice conveys.
As you make the adjustments, Siri will practice saying some common things, like “You have one new message,” so you can get a sense of how the different voices sound.
The AI version of Siri is deeply integrated across the updated version of iOS, where it will allow iPhone owners to start conversations by speaking, swiping down from the Dynamic Island at the top of the screen and typing, tapping on the phone’s side button, or even by using the brand-new stand-alone Siri app.
Other, more minor updates are also rolling out with iOS 27 beta 3, including an updated Reminders app icon. (We should note some people on X are also reporting losing access to the new Siri after updating, or seeing their phone again begin indexing their data — typically, the first step in optimizing Siri AI for search.)
Utah has become the first US state to let an AI chatbot, Doctronic, renew prescriptions without a doctor, via a regulatory sandbox that waives licensing laws. The state’s medical licensing board, blindsided by the January launch, called in April for the pilot to be halted over safety risks, but the state refused. The case exposes a federal-state regulatory vacuum around AI in medicine.
TL;DR
Utah has quietly become the first US state to let an AI chatbot renew prescriptions without a doctor, according to the Associated Press. The programme, run by a company called Doctronic, launched in January and has set off a fierce medical debate.
Residents can skip the doctor’s office and refill prescriptions online through the chatbot. It asks about their medication and history, checks a national pharmacy database, and either renews the script or escalates to a human doctor.
The launch was possible only through a “regulatory sandbox” that lets Utah officials waive laws for promising AI. State and federal rules otherwise restrict prescribing to licensed medical professionals.
“We have crossed a threshold in terms of giving something that is not human a medical license, whether or not we want to call it that,” the University of Pennsylvania’s Dr Eric Bressman told the AP. He and others say they are not opposed to AI prescribing, but want it held to standards as rigorous as those for human doctors.
Utah’s medical licensing board says it only learned of the programme when the January launch made the news. In an April letter, 11 members called for the pilot to be halted, citing the risks of auto-renewing drugs with side effects or interactions.
“We were essentially told: ‘Yes this is going on. And no, you don’t have a say in it’,” said Dr Alan Smith, a family physician who chairs the board but spoke for himself.
The state declined to suspend it, noting human doctors still review every refill in this first phase.
The programme is currently overseen by a five-member board of AI specialists, none of them doctors. Doctronic expects to move to fully automated refills soon.
Smith warns the risks are real, pointing out that Doctronic’s roughly 190 refillable medications include blood thinners, which turn dangerous if a patient develops internal bleeding. The American Medical Association has echoed the concern that “prescription renewals aren’t routine checkboxes”.
The case exposes a jurisdictional tangle, since medical technology is regulated federally while medical professionals are overseen by states. Doctronic frames its AI as part of state-regulated medical practice, though some experts argue it has crossed into FDA territory.
The company would not say whether it has sought FDA permission. The agency told the AP it has authorised no AI chatbots but wants to encourage innovation, a hands-off posture that fits a broader loosening of oversight on AI health tools.
Critics see history rhyming, with Bressman comparing the moment to the haphazard medicine of the early 20th century, before boards and benchmarks existed. The template for licensing AI medical services in other states comes from the Cicero Institute, a pro-AI think tank founded by Palantir co-founder Joe Lonsdale.
The stakes are not abstract, as safety researchers have warned that medical chatbots can sound authoritative while dispensing dangerous advice. Others caution that removing humans from care can undermine the very outcomes it promises.
Rivals are scrambling to map those failure modes too. Meta went as far as posing as teenagers to test how competing chatbots handle sensitive topics.
Doctronic plans peer-reviewed studies later this year, though its only published paper so far was written by its own scientists and not independently reviewed. As one Utah law professor put it, companies risk letting the technology race beyond the evidence, and betraying public trust in the process.
US District Judge Charles Breyer denied Elon Musk’s bid to overturn a March 2026 jury verdict finding he defrauded Twitter investors during his 2022 takeover, upholding the finding on his 13 May bot tweet while granting one narrow point on a 17 May tweet. Investors say damages could reach $2.6bn, and the judge also granted prejudgment interest.
TL;DR
A federal judge has refused to overturn a jury’s finding that Elon Musk defrauded Twitter investors during his $44bn takeover of the platform in 2022. US District Judge Charles Breyer denied Musk’s motion to set aside the verdict in most respects on Monday.
A San Francisco jury ruled in March that two of Musk’s May 2022 tweets about the deal and Twitter’s spam bot numbers were materially false or misleading. Investors say the resulting losses could support damages of up to $2.6bn.
“Buyer’s remorse is not an exception to the securities laws,” Breyer wrote, adding that the laws are “in their essence, about trust”. The judge found substantial evidence that Musk’s 13 May tweet, claiming the deal was on hold pending bot data, was literally untrue.
Breyer cited testimony from one of Musk’s own bankers, who said the tweet surprised her and that Musk never actually put the deal on hold. A jury could infer Musk had a motive to escape the deal and used bots as a pretext, the judge wrote.
He did hand Musk one narrow win, agreeing there was too little evidence that a separate 17 May tweet caused investors a market loss. Musk’s lawyers did not immediately respond to requests for comment.
The case traces back to Musk’s chaotic pursuit of Twitter, when he agreed to buy the company, then tried to walk away citing spam accounts. Twitter sued to force the deal through, and Musk ultimately closed at $54.20 a share before renaming the platform X.
Investors sued in October 2022, arguing Musk deliberately talked the stock down to renegotiate or exit. The jury agreed he misled the market, though it rejected the broader claim that he ran a deliberate scheme.
Breyer also swatted down Musk’s more colourful arguments, including a claim that jurors mocked him by writing “$4.20” in blue ink on the verdict form. The number references cannabis, the judge noted, and the jury had actually cleared Musk on two claims.
The ruling adds to a crowded docket for Musk, who recently settled a separate SEC case over his late disclosure of an initial Twitter stake for $1.5m. His “funding secured” Tesla saga first drew SEC fraud charges back in 2018.
He is also fighting Sam Altman in a high-stakes trial over OpenAI, all while steering the newly public SpaceX. The tweets that built his mythology keep generating legal bills.
Prejudgment interest, which Breyer also granted, could push the final figure higher still. For a man now worth more than a trillion dollars, the sum is survivable, but the finding that he defrauded investors is harder to shrug off.
Weekend Open Thread: High Hopes
The House | “Reframing the debate from a binary discussion of winners and losers”: Yuan Yang reviews ‘We Are Not Machines’
Strategy authorizes up to $1.25B in Bitcoin sales under new capital plan
Open Thread: What Great Books Have You Read Recently?
How to Build INSANE Live Financial Dashboards With Claude
Taylor Swift and Travis Kelce wedding staffer hilariously struggles to keep her cool while checking in megastars
Anonymous researcher drops 0-day ‘exploitarium’ repo
Australia treasurer says alleged access of prime minister’s bank data ’incredibly concerning’
Airdrop Registration Becomes Key Focus For Remittix As RTX Launch Updates Approach
Broncos roster: OL Ben Powers (No. 74) entering final year of contract
Binance stock trading tops $1B in first month after launch
Presenter Caroline Flack’s brother Paul Flack dies aged 55
Alibaba-affiliate Ant Group enters the humanoid robot market with 12 deals
Standard Chartered Secures MiCA License as ESMA Adds 37 New Crypto Firms
South Africa proposes crypto tax guidance under existing rules
Best Time to Enter Small Caps Right Now? Another Bull Run? | Financially Free
New exhibition reflects five decades of movement between island of Ireland and GB
Meta Platforms Stock Jumps 7% Today as Bloomberg Reports Company Plans to Enter the Cloud Business
Lenovo laptops are now shipping with YMTC SSDs, a sign of Chinese NAND entering the mainstream
What a 10 Percent Drop Means for Buyers, Sellers and Renters
You must be logged in to post a comment Login