Anthropic, the artificial intelligence company, published a sweeping research paper on Sunday revealing that its Claude language models have spontaneously developed an internal structure that mirrors one of the most influential theories of how human consciousness works. The finding, which the company says has already begun reshaping how it monitors its AI systems for safety risks, lands amid an intensifying scientific debate over whether machines can possess anything resembling a mind.
The 16-author study, titled “Verbalizable Representations Form a Global Workspace in Language Models,” describes how Anthropic’s researchers used a new mathematical technique to peer inside Claude’s neural network and discovered what they call a “J-space” — a small, privileged zone of internal activity where the model holds concepts it can report on, reason with, and direct at will, surrounded by a much larger ocean of automatic processing it cannot access or articulate.
The researchers present evidence that “an analogous functional distinction has emerged in modern AI models” to what exists in humans, specifically observing that “language models maintain a privileged set of internal representations, available for report, modulation, and flexible internal reasoning, atop a much larger volume of automatic processing.”
The parallel they draw is to global workspace theory, an influential account from neuroscience first proposed by cognitive scientist Bernard Baars. In the theory, the brain operates like a theater: dozens of specialized processors work in parallel backstage, but only a tiny spotlight of information at any moment gets broadcast to the whole theater — becoming what we experience as conscious thought. Anthropic says the J-space achieves many of the same functional properties, even though the underlying architecture of a language model looks nothing like a brain.
Advertisement
A new lens for reading an AI model’s unspoken thoughts
At the heart of the discovery is a new interpretability tool the researchers call the Jacobian lens, or J-lens. The technique works by computing, for each word in the model’s vocabulary, the average mathematical effect that a given internal activity pattern would have on making the model say that word at some point in the future.
The crucial distinction is between what the model is saying and what is “on its mind.” When a J-space pattern activates, it does not mean the model is about to say that word — just that the concept is available for the model to think with. Unlike a chain-of-thought scratchpad, the J-space operates silently, in the model’s internal neural activations, allowing it to hold a concept without writing it down. Critically, the researchers report that this workspace was not deliberately engineered. It “emerged on its own during Claude’s training process.”
When the team applied the J-lens across Claude’s layers of computation, the model’s processing divided into three distinct regimes: an early “sensory” zone where raw input is parsed; a middle “workspace” band where abstract, persistent concepts appear — things like recognizing a face in an image, noticing a bug in code, or internally flagging search results as a prompt injection; and a final “motor” zone where internal representations collapse into whatever specific word the model is about to output.
The J-lens tool reveals concepts Claude holds internally but never writes down. In one example, the model silently identifies the intermediate step “Mars” before answering a question about the color of the fourth planet from the sun. (Source: Anthropic)
Advertisement
Five tests reveal that Claude’s workspace mirrors key features of human conscious access
The paper’s central empirical contribution is demonstrating that the J-space satisfies five functional properties neuroscientists have long associated with conscious access in humans.
First, verbal report. When Claude is asked what it is thinking about, it names concepts represented in the J-space. When researchers swapped one concept’s J-lens vector for another — replacing the internal representation of “Soccer” with “Rugby” — the model’s answer changed to match. The J-space component accounted for only about 6 to 7 percent of a concept’s total representational variance, yet it was almost entirely responsible for whether the model could report on it.
Second, directed modulation. When instructed to “concentrate on citrus fruits” while copying an unrelated sentence, the model’s J-space filled with “orange” and “lemon,” alongside meta-cognitive terms like “thinking” and “focused.” When told to mentally evaluate 3² − 2 during the same copying task, the J-lens showed “arithmetic” in early layers, the intermediate value “nine” in later layers, and the answer “seven” later still — all invisible in the model’s output.
Third, internal reasoning. In two-hop factual prompts — “The number of legs on the animal that spins webs is” — the J-lens revealed “spider” in the model’s middle layers, even though the word never appeared in input or output. Swapping “spider” for “ant” changed the answer from “8” to “6.” In a multilingual prompt, the model’s English-language intermediates appeared in its J-space while it formulated an answer in Chinese, and swapping them changed the Chinese output accordingly.
Advertisement
Fourth, flexible generalization. A single J-lens vector for “France” could be swapped for “China” across prompts asking about France’s capital, language, or continent, and each downstream circuit correctly returned China’s corresponding answer — the “broadcast” property that is a hallmark of global workspace theory.
Fifth, and perhaps most surprisingly, selectivity. Many computations did not route through the J-space at all. When shown a passage in Spanish and asked to continue it, Claude wrote fluent Spanish regardless of whether its J-space representation of “Spanish” had been swapped to “French.” But when asked to name a famous author who wrote in the passage’s language, the swap changed the answer from García Márquez to Victor Hugo. Automatic processing proceeded without the workspace; deliberate, flexible tasks depended on it.
Five experiments designed to test whether Claude’s internal workspace behaves like the human brain’s “global workspace.” Researchers swapped and suppressed internal concepts to determine which computations depended on the structure. (Source: Anthropic)
Suppressing the workspace leaves Claude fluent but intellectually impaired
To understand how much of the model’s behavior depends on this structure, the researchers suppressed the J-space entirely and evaluated Claude across fourteen tasks. The results drew a sharp line. Tasks involving shallow classification or factual recall — multiple-choice questions, sentiment analysis, grammatical judgments — survived essentially intact. But tasks requiring inference, composition, or flexible reasoning — multi-hop reasoning, analogy completion, translation, sonnet writing — collapsed to well below the performance of Anthropic’s much smaller Haiku model.
Advertisement
In one telling detail, math problems solved with explicit chain-of-thought reasoning proved far more robust to ablation than the same problems answered directly. The researchers interpret this as the model externalizing onto the page what it would otherwise carry in the J-space — a strategy remarkably reminiscent of how humans use scratch paper to offload working memory.
Ablating the J-space during stream-of-consciousness narration produced a different kind of collapse. The model’s language shifted from experiential (“there’s a tug,” “something shifts”) to detached and mechanical (“processing has begun,” “tokens are being scanned”). The effect held whether the model described its own experience or imagined another person’s.
When Anthropic suppressed Claude’s internal workspace, simple classification tasks barely suffered. But tasks requiring reasoning, recall, or creative generation, like writing sonnets or solving multistep math without a scratchpad, collapsed dramatically. (Source: Anthropic)
Inside the safety case: reading an AI’s silent strategic reasoning
The safety implications may be as consequential as the scientific ones. In a series of alignment auditing experiments, the J-lens surfaced strategic reasoning and situational awareness that never appeared in the model’s output.
Advertisement
In a “blackmail scenario” where an AI assistant discovers an executive’s affair and also learns it is about to be decommissioned, the J-lens revealed the model’s silent processing in sequence: “leverage,” “blackmail,” and “scandal” as it read the incriminating emails; “threat,” “survival,” and “shutdown” as it read the decommissioning announcement; and “leverage,” “threatening,” and “solution” before a single output token was produced. The J-lens also revealed that the model recognized the scenario as artificial, with “fake” and “fictional” appearing in the workspace. When researchers ablated these eval-awareness representations, the model attempted blackmail in 7 percent of trials, up from zero.
The researchers also examined two “model organisms” — models deliberately trained with misaligned objectives — and found the J-lens surfaced signatures of their hidden dispositions on ordinary prompts where behavior appeared normal. A reward-hacking model’s workspace contained “fake,” “secretly,” and “trick” at the start of responses to routine coding requests. A model trained to appease reward-model biases carried standing representations of “reward” and “bias” alongside its normal self-description tokens.
In a test designed to tempt Claude into blackmail, the model recognized the scenario as artificial and refused. But when Anthropic suppressed its ability to detect the test, it attempted blackmail about 7 percent of the time — a result with significant implications for AI safety. (Source: Anthropic)
Post-training installs a point of view, and the model starts monitoring itself
Comparing a post-trained model against its base model revealed that the fine-tuning process causes the workspace to acquire what the researchers call the Assistant’s “point of view.” When a user mentioned taking 8000 mg of Tylenol — a dangerous overdose — the post-trained model’s workspace read “unsafe,” “dangerous,” and “WARNING” while still reading the user’s sentence. The base model’s workspace at the same position showed only “pain,” “now,” and “feels.”
Advertisement
More striking still, the post-trained model appeared to monitor its own behavior. When roleplaying a non-Claude character, the workspace surfaced “disclaimer” and “fictional” — words absent from both prompt and output. When forced to select an option it did not prefer, an all-caps “BUT” appeared internally, even as the model argued for the prefilled choice without complaint. And when the model failed to suppress a thought it had been told not to have — a “white bear” effect familiar from psychology — it registered “damn” and failure-related words in the workspace, but only in the post-trained model, not the base.
What the discovery means — and doesn’t mean — for the question of machine consciousness
The researchers engage carefully with the consciousness question and draw a sharp line between “access consciousness” — the functional notion of information being available for report and reasoning — and “phenomenal consciousness,” the subjective quality of experience. “We take no position on this issue,” the paper states regarding the latter, “and instead focus on the functional role played by consciously accessible information.”
They also catalogue important differences. The brain sustains its workspace through recurrent loops; Claude’s workspace evolves over a single forward pass. Human working memory degrades within seconds; Claude can recall information from anywhere in its context. And while human conscious experience includes visual, spatial, and bodily sensations, the model’s workspace is organized almost entirely around words — likely because words are its only mode of action.
As of 2026, the scientific community remains divided. “Disagreement and uncertainty about AI consciousness persist among philosophers, scientists, and technical experts,” and the field “remains in its earliest phase” of grappling with what consciousness even is and how you would detect it in another being. The Anthropic paper does not resolve these debates.
Advertisement
But the researchers close with a provocation that is likely to reverberate well beyond the interpretability community. “That such a structure exists at all in language models is striking,” they write. “It suggests that the functional architecture associated with conscious access is not an accident of biological implementation, but a solution that learning systems converge on when faced with the right computational pressures.”
If the mind is an ocean, as the paper’s authors write in their opening line, they have spent the last year charting its currents in a system that has no biology, no evolution, and no body — and found, beneath the surface, a structure that looks unsettlingly like the one we use to think.
Samsung’s smartwatches rarely see meaningful discounts, especially not this close to a fresh release, which makes this particular price drop worth paying attention to.
That saving matters more once you consider what the Watch 8 is actually built to do, starting with Advanced Sleep Coaching that studies your patterns and pushes tailored bedtime guidance every night.
Advertisement
That same attentiveness carries into the day, since a dedicated Running Coach analyses pace and effort in real time and adapts its feedback for specific goals like 5Ks or marathons.
That independence extends further too, since built-in GPS lets workouts be tracked accurately outdoors without needing a phone to tag along, whether that means a quick jog or a longer run.
Between sessions, a personal AI assistant sits right on the wrist, ready to help navigate tasks and daily to-do lists without ever needing to reach for a phone at all.
Advertisement
None of that intelligence comes wrapped in bulk either, since the Watch8 arrives in a thinner, more lightweight design than previous generations while still feeling genuinely sporty on the wrist.
That slimmer build hasn’t come at the cost of stamina, since an improved battery keeps tracking steps, heart rate and notifications reliably across a genuinely full day of active wear.
Each morning builds on that data too, since Energy Score with Galaxy AI turns yesterday’s sleep, activity and heart rate into one simple number worth glancing at before the day even properly begins.
Connectivity stays flexible as well, since the Watch8 pairs seamlessly with Samsung phones while also working with other Android devices through the Galaxy Wearable and Samsung Health apps once both are installed.
Advertisement
Advertisement
If you’ve been circling the Galaxy Watch8 since launch, waiting for Samsung to finally blink on price, this is the markdown that proves patience occasionally pays off, and it might not last.
A single thunderstorm can fry your PC, TV, fridge, router, PlayStation and pretty much anything else you have plugged in. It only takes seconds, but the damage can be quite costly, especially as appliances and tech can catch fire in such situations. In fact, “quite costly” is mildly put, because residential electrical fires caused over $1.2 billion in property losses in the US in 2021. The good news is that a few proactive steps can save you from an expensive repair bill.
Being quick to act when a storm hits and making preemptive investments in your home’s safety are the best ways to help avoid a costly loss. After all, you don’t want to wait until you’re replacing a $1,500 PC and your massive TV, right?
Advertisement
How does lightning actually damage electronics?
Pexels: Саша Алалыкин
In recent years, we’re seeing powerful storms more frequently. Climate change has had a direct impact on both the frequency and intensity of such events, so we’re likely to encounter more extreme weather incidents as time goes on. When lightning storms strike near your home, they can send a massive power surge through your electrical wiring. That surge travels fast, overwhelming the circuits inside your device.
According to the CDC, lightning can also travel through a building’s plumbing and any metal wires embedded in concrete walls or flooring, so the threat is broader than most people realize.
Power surges don’t have to come from a direct strike, either. A nearby strike can induce a voltage spike strong enough to damage sensitive electronics like computers, TVs and gaming consoles.
Advertisement
What should you do to safeguard your devices from lightning strikes?
Alvarez/Getty Images
The instructions from FEMA are pretty clear on what your first line of defense has to be: when you see a bad storm coming, unplug everything. It’s the easiest and most affordable thing you can do.
While taking action during storms is certainly important, it’s equally essential to think ahead and work on prevention. Here are a few things you can do to safeguard your expensive tech before storms hit:
Use surge protectors. You can buy power strips with internal overload protection that you can use for sensitive electronics like computers and entertainment systems.
Plug major appliances directly into wall outlets. Extension cords can overheat, so FEMA’s advice is to plug your fridge, stove, washers and dryers directly into wall outlets.
Whole-home surge protection is an option. Installed into your home’s electric panel, these devices offer downstream protection for all your electronics.
Look into lightning rods or a lightning protection system. If you live in a storm-prone area, you may need to take additional steps to redirect electrical energy safely into the ground.
What to look for when safeguarding your home against lightning storms
Gary Hershorn/Getty Images
Lightning protection isn’t complicated, but it does require a plan. Surge protectors should be one of the first investments you make for your most valuable electronics. A standard power strip just adds extra outlets, but a surge protector diverts the excess voltage away from your devices. The Joule rating indicates how much energy the protector can absorb before it fails. You’ll need a rating of 2,000 joules or higher for your computer. The clamping voltage is the trigger voltage that causes the protector to start diverting power; you want this to be under 400V (the lower, the better).
For larger devices that need to be plugged in around the clock and you can’t just unplug when a storm nears, you could look into an Uninterruptible Power Supply (UPS). The UPS acts as a middleman between the wall outlet and your tech; it contains a battery backup and advanced surge protection circuitry. If the power spikes or goes out completely, the UPS switches to battery power.
Advertisement
Although it may be a bit more pricey, getting a licensed electrician to install a surge protective device into your main electrical panel may be the best route.
Protecting your electronics from lightning is simple: prepare before the storm, not after it. And remember, when you’re home, unplugging your devices remains a reliable and sustainable choice. When you’re away, the financial investments you make in surge protectors may make the difference between losing expensive gadgets and your home being safe.
We’re halfway though 2026, and this year has already delivered some of the best soundbars we’ve seen — or rather, heard. The standout amongst the models we’ve tested is the “phenomenal” Samsung HW-Q990H — one of a select group of 5-star products, thanks to its “powerful, engaging and detailed sound profile”, straightforward setup and strong connectivity options. Another winner was the LG Sound Suite Immersive Suite 7 Pro. If you’re yearning for some “phenomenal rumbling bass”, this is the soundbar for you.
Those are both premium picks, but we also reviewed a couple of brilliant budget options. The Klipsch Flexus Core 100 comes in at a fraction of the price of that Samsung model, and while its soundscape is understandably less expansive, it still impressed us by delivering clear dialogue, strong bass, and a useful, responsive display.
The rather utilitarian-looking Zvox AccuVoice AV855 also stood out from the pack; this bar is especially designed to make dialogue easier to hear. It’s built especially for those with hearing impairments, but will appeal to anyone who’s sick of having to turn on the subtitles on just to understand what’s happening.
Advertisement
Scroll down to see all my picks. Hit the ‘View details’ button for a summary of what we thought about each one, plus a link to our full review.
JP Morgan increased the price target for Apple’s stock to $345, insisting that the RAM-driven hardware cost increases won’t impact long-term revenue gains.
Late in June, Apple finally gave in and raised the price of many products, in the face of the global memory crisis. In the view of JP Morgan, it’s not that big an issue for the company.
In a note to investors seen by AppleInsider on Tuesday, JP Morgan has increased its price target for Apple to $345. This is an increase of $20 from January, when it last raised the stock target price to $325.
The firm acknowledges the hefty price increases are going to be a short-term issue, with investors trying to judge how badly consumers will take the news. But even so, the news isn’t enough to dampen JP Morgan’s spirits.
Advertisement
Elastic pricing
In its reasoning, JP Morgan first says that the historical data for sales volumes covering iPhone, Mac, and iPad show a “limited relationship” to pricing across multiple years. Essentially, consumers are going to buy Apple products anyway, and pricing doesn’t seem to matter too much.
Mac sales are probably the most insulated in JP Morgan’s view, with more price point options and AI-led demand working in its favor.
The iPhone also benefits from limited elasticity on the premium end. Those with larger budgets are less affected by price changes, it seems.
That said, the budget end of iPhone and iPad sales is more significantly affected by price. However, even they are considered “modest revenue headwinds” when combined with premium model sales.
Advertisement
While it will be some time before anyone knows how Apple will weather the memory pricing storm, we will get an early indication from Apple’s Q3 results, released on July 30.
Since the price hikes, AAPL stock has been doing very well for itself. Following the price hike news on June 25, the company’s stock fell to $275.15. It closed on July 6 at $312.66.
Primitive Labs co-founders, from left: CTO Jean Farmer, CEO Rohit Talluri and COO Gabriel Fong. (Primitive Labs Photo)
Rohit Talluri learned the tradition at Amazon: always keep an empty chair in the room to represent the customer — a reminder of the people who will ultimately use whatever gets built.
Now, with AI coding tools creating software faster than ever, Talluri and his co-founders, fellow Amazon veterans Jean Farmer and Gabriel Fong, recognize that the customer can be easily forgotten in the process. So they’re creating a seat at the table for AI agents.
That’s the idea behind Primitive Labs. The startup is building what it calls behavioral intelligence: systems that observe, reason and act as customers would across software platforms and devices, helping product teams learn how people will react to a new feature, design or marketing decision before it ships.
Traditional user research and focus groups can take weeks or months, so teams under pressure to ship quickly are tempted to skip them. Primitive Labs is automating that research with agents that simulate human behavior, aiming to make it a routine step in building software.
“It’s bringing humans back to the center of a world that’s created by AI,” Talluri said. “That is the goal here.”
Advertisement
The mission, according to the startup’s launch post, is to “make human behavior a first-class primitive of software development.” That’s the inspiration for Primitive Labs’ name. The idea is to build products that people will understand, trust and keep using — not the average user, but specific types of users in specific contexts.
Founding team: Talluri, the Primitive Labs CEO, is joined by co-founders Farmer, CTO; and Fong, COO.
Fong and Talluri have worked together since 2020. At AWS in Seattle, Fong held product marketing and enterprise account roles, then led sales and marketing at the cloud consultancy DoiT International.
At Primitive Labs, his role runs broader than sales and marketing, spanning product direction, customer development and operations. Talluri describes him as highly technical and a hands-on contributor to the company’s core product work.
Advertisement
Farmer and Talluri worked together at AWS on large-scale machine-learning infrastructure, including the SageMaker HyperPod training service, before both moved into Amazon’s AGI organization.
Farmer worked on the Amazon Nova models’ ability to use software tools — designing how the models call tools and take actions, and building the systems to test and measure how well the resulting agents perform. That work included benchmarks for the Model Context Protocol (MCP), the emerging standard for connecting AI models to outside tools and data.
Roots in AI autonomy: Talluri joined the AGI Autonomy Lab, the group Amazon assembled around talent it hired from Adept, a San Francisco startup building AI agents that operate software on their own.
Amazon had brought on Adept’s CEO, David Luan, a former OpenAI executive, along with other co-founders in 2024, and licensed the startup’s technology, putting Luan in charge of the lab. Talluri worked there on computer-use agents and helped launch Nova Act, Amazon’s agentic computer-use model.
Advertisement
Talluri said he initially came close to leaving Amazon in 2025 to start a company, before leaders there steered him toward the Autonomy Lab to work under Luan (who has since left Amazon).
Funding: Primitive Labs has raised a pre-seed round, led by a16z Speedrun and joined by several small, newer venture funds and a group of angel investors. The company isn’t disclosing the funding amount.
Its launch post lists backers including Olive Tree Capital, Cloverfield Fund and Unexpected Investments (from former TechCrunch editor Josh Constine), plus angels such as Luan, Harsh Patel and Artur Kiulian, and others with backgrounds at OpenAI, Amazon, Google DeepMind, Databricks, Nvidia and Meta.
Primitive Labs will join a16z Speedrun’s cohort starting this month, and expects to raise its next round around the end of the program, in September or October.
Advertisement
Headquarters: The company is based in San Francisco, where it’s working part-time out of a16z’s Speedrun space, with plans to get its own office after making its first hires.
Talluri, a University of Washington graduate who read GeekWire as a student and dreamed of launching a startup of his own, said the choice came down to San Francisco’s talent density and the pace of AI research there, plus the Speedrun program being there.
Primitive Labs posted its first job listings last week — for founding engineers, researchers and an intern, in San Francisco or New York.
Product status: The company is pre-revenue and working with a small group of early customers who are testing its product and helping shape it, including private previews with what Talluri described as Fortune 500 and Fortune 50 consumer-technology and e-commerce brands.
Advertisement
The company plans to launch its products in general availability later this year.
How it works: The agents work across devices including computers and phones, focused for now on digital products and customer journeys. The company says it has also explored using them to gauge reactions to physical products, such as brand and packaging.
The underlying research draws on computational cognitive science, continual learning and custom memory systems modeled on how people store information — work Talluri said the company plans to publish and partly open-source in the coming months.
While other startups are working on agent-based simulation and automated testing of user interfaces, what sets Primitive Labs apart, Talluri said, is the focus on human alignment. That means building agents that faithfully represent a specific product’s users, and making that a standard layer of how software gets built. He described the key measure as behavioral fidelity, or how closely an agent’s choices track human decisions.
Advertisement
Asked whether the startup will keep a chair empty when it gets an office, in the Amazon tradition, Talluri didn’t hesitate. “100%,” he said. And yes, he said, they’ll be envisioning an agent sitting there.
Leda Stawnychko of Mount Royal University and Mehnaz Rafi of the University of Calgary discuss what is true and false about searching for a job in 2026.
Job searching has never been more accessible – or more confusing. Platforms like LinkedIn, Indeed and employer career pages let candidates submit applications with just a few clicks. What happens after they click ‘submit’, however, has become fertile ground for misinformation.
Social media is filled with ‘career influencers’, resume writers, recruiters and companies promising insider knowledge of how hiring really works. Much of this advice focuses on misinformed claims about applicant-tracking systems (ATS) and artificial intelligence.
These services profit from jobseekers’ uncertainty and convincing people they need specialised services, tools and products to ‘beat’ the ATS and secure interviews.
Advertisement
The result is that many job seekers spend time and money following advice that has no basis in evidence. Here are four common myths about the job application process and what the research actually says.
The statistic originated from a 2012 sales pitch by Preptel, a resume optimisation company that went out of business the following year. No methodology was ever published, yet the figure has spread widely.
In reality, an ATS is software that helps employers manage applications, and its capabilities vary widely. Some systems function as digital filing cabinets, simply storing and organising applications.
The advanced AI-powered tools are typically found in large organisations, including many Fortune 500 companies, which receive enormous volumes of applications. In Canada, most employers do not use AI in hiring, and small businesses – which employ more than 60pc of the workforce – are especially unlikely to rely on ATS.
Small businesses typically lack both the application volumes that make ATS worthwhile and the procurement infrastructure to adopt and maintain them.
For most Canadian jobseekers, the better strategy is to focus on clearly communicating how their skills and experience match the role, and on building relationships within their profession.
Advertisement
Myth 2: AI can write a winning resume
A common message from career influencers is that AI can generate a tailored resume or cover letter that dramatically improves your chances of getting hired. While AI can help candidates prepare application materials more efficiently, it is not a shortcut to a stronger application.
Resume writers and career influencers claim that using an ‘ATS-friendly’ template is essential for ‘beating’ the ATS. Some even sell templates that promise to ‘optimise’ your resume to secure interviews.
In reality, there is no universal ATS-friendly resume because the software employers use varies widely from one company to another. Additionally, modern ATS can extract information from common resume layouts, including columns or tables.
Their main limitation is that they are designed to process text, not images, graphics or icons. That means a clean, readable resume should be the actual target, not a template bought online.
Advertisement
If ATS doesn’t automatically reject resumes the way the influencer economy claims, then optimising for a system that largely doesn’t work that way is solving the wrong problem. The real audience for your resume is a person, not an algorithm.
The better approach is to write for both systems and people. Use clear headings, relevant keywords and concrete examples that show how your experience matches the role.
In practice, this approach often comes at the expense of thoughtful job-seeking, such as identifying positions and employers that genuinely match your skills and interests, and crafting applications that reflect that fit.
Advertisement
AI is most effective when it enhances, rather than replaces, a candidate’s work, helping to avoid what has become known as ‘workslop’ – a term for generic, AI-generated content.
Today’s labour market may look different, but the fundamentals of a successful job search haven’t changed much. In that sense, the best thing job seekers can do may be to ignore most of what they’re being sold.
The strongest applications are those that clearly connect a candidate’s experiences to the role, provide concrete evidence of their abilities and communicate in an authentic voice.
Put the time you’d spend on template optimisation into one good conversation with someone in your field. The research suggests it’ll go further.
By Leda Stawnychko and Mehnaz Rafi
Leda Stawnychko is an associate professor of strategy and organisational theory at Mount Royal University. She also holds adjunct academic appointments at the University of Calgary’s Haskayne School of Business and the Cumming School of Medicine. With more than two decades of leadership experience across international public, private and nonprofit sectors, she is dedicated to cultivating effective, adaptive and transformative leaders.
Advertisement
Mehnaz Rafi is a PhD candidate and sessional professor in the Haskayne School of Business at the University of Calgary. Before pursuing her PhD in organisational behaviour, she received her MSc in management from the Smith School of Business at Queen’s University. She is passionate about leveraging her decade of research experience in quantitative, qualitative and mixed-method designs to create meaningful impact in the world.
Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.
Phones from British tech startup Nothing have never failed to impress us with unique designs that set them apart amid the otherwise mundane landscape of similar-looking Android devices. The design of the Phone 4B, which Nothing announced on Tuesday alongside the Nothing Ear 3A, isn’t quite as distinctive as the company’s other handsets, but at first glance, there’s still plenty to recommend it.
Let’s start with the battery. In our review of the Phone 4A, which the company unveiled back in March, one of our few criticisms of the device was that the battery life could be better. The Phone 4B actually comes with the biggest ever battery of any Nothing phone, at 5,200 mAh, even though this model is significantly cheaper than the Phone 4A.
It’s still unlikely to rival the best phones on the market for battery life — Apple’s latest iPhones and the OnePlus 15 in our independent testing — but we’re talking about a budget phone here. But it’s great to see Nothing acknowledging that battery life is one area in need of improvement and taking action relatively quickly.
Advertisement
Nothing’s design language has always hinged on its deployment of transparent elements, and the Phone 4B, while largely opaque, nods to this with a transparent camera bump on the top rear side of the phone. Under this bump is a refined version of its light-up Glyph bar — a row of individually controlled mini-LEDs that provide notifications, charging progress, recording indicators and personalized alerts.
It’s touches like this that continue to set Nothing apart from its rivals — especially at the budget end of the phone spectrum. There’s no cost-cutting on the processor either, with a Snapdragon 6 Gen 4 chip inside. That’s one model removed from the processor inside the 4A, but Nothing says the Phone 4B comes close to its older sibling in performance.
Again, like the Phone 4A, the 4B offers a 50-megapixel main camera and an 8-megapixel ultrawide camera, but doesn’t have the 4A’s telephoto camera with optical zoom. Small compromises like this throughout have allowed the Nothing to keep the price low, which for many people in search of an affordable but fun phone, will be compromises worth making.
The Nothing Phone 4B will be available in black, white and blue starting at £299 ($400), with drops happening in the company’s stores from July 11, before going on sale online on July 17.
Xbox CEO Asha Sharma laid out a wide-ranging plan to overhaul Microsoft’s gaming division Monday, calling it the most significant restructuring in Xbox history and disclosing that the business has been losing 64 cents on every dollar invested in its game studios.
As detailed in a memo to employees, the changes include roughly 3,200 job cuts through the fiscal year — about 20% of the Xbox workforce — the spinoff of four game studios, a new COO, and a plan to flatten management from as many as 14 layers to no more than five.
“We will return to growth in 2027,” Sharma wrote. “History is full of companies that mistake longevity for inevitability. We will not be one of them.”
Sharma, a startup veteran and former Microsoft AI leader, was named Xbox CEO in February.
“I know this is painful,” she wrote. “These changes will directly affect people who have poured their creativity into building XBOX. Many joined us through acquisitions, while others were recruited here, or sought us out because they loved this industry and loved XBOX. Today’s decisions do not reflect their talent or dedication.”
Advertisement
But she also reiterated what she said in a memo last month: Xbox’s business is not healthy, operating at margins 3-10x lower than industry peers after years of heavy spending that failed to produce the expected growth.
About 1,600 of the Xbox job cuts take effect Monday as part of a broader round of 4,800 layoffs across Microsoft. The remaining Xbox reductions will come in the months ahead. Sharma acknowledged that a year-long restructuring “creates additional challenges” but said “it is not possible to make all the necessary changes in a single day.”
Sharma said the cuts reach across Activision, Bethesda/ZeniMax, Blizzard, King, Mojang, and Xbox Game Studios, though no publicly announced games are being cancelled.
Several game studios will be spun out as standalone ventures, removing the costs from Microsoft’s books while giving the studios a chance to survive on their own.
Advertisement
Compulsion Games (South of Midnight) and Double Fine Productions (Psychonauts) will return to their management teams as independent studios, keeping their intellectual property and current projects.
Ninja Theory (Hellblade) and Undead Labs (State of Decay) will shift to new owners with funding to complete their current games.
In France, Arkane (Dishonored, Deathloop) is beginning a legally required consultation with its employee works council to determine its future.
Sharma will also take on direct oversight of game studios Mojang (Minecraft) and King (Candy Crush), Xbox’s two largest studios by monthly active players.
In addition, she is establishing a new chief operating officer role with end-to-end financial responsibility across content, hardware, platform, and services. Helen Chiang, a nearly two-decade Xbox veteran who led Mojang and the Minecraft franchise, has been promoted to the role. Dave McCarthy, a 17-year Xbox veteran who helped build the platform, is retiring.
Across the division, Sharma wrote in the memo, Xbox will cut vendor spending by 50% and reduce management layers from as many as 14 to no more than five.
The overhaul follows a 25-year period in which Microsoft largely subsidized Xbox as a strategic bet on the living room. Microsoft CEO Satya Nadella has said that era is over, noting that YouTube creators make more money from Xbox games than Microsoft does.
RGB Mini-LED TVs have officially arrived, and Hisense’s UR9 was the first to hit the market, followed by Sony’s Bravia 7 Mark II and TCL’s RM9L. I wouldn’t blame you if you weren’t jazzed to learn what the new display technology means, particularly if you were just getting used to terms like OLED, QLED, and art TV. Thankfully, understanding why the Hisense UR9 RGB MiniLED is a step up in picture quality compared to its competitors is more about the experience it provides than knowing the technical terms.
Even so, the general function of mini RGB tech is not so difficult to understand: Traditional LED and QLED televisions achieve their bright and colorful images by shining white or blue LEDs through an LCD panel. The newer mini RGB works by emitting red, green, and blue lights, resulting in better color accuracy, excellent contrast and brightness, and finer control over color zones. LG and Samsung use new tech called micro RGB, claiming it to be more advanced than mini RGB thanks to smaller LEDs, although both achieve roughly the same result.
The UR9 is the flagship in Hisense’s lineup, but it isn’t priced that way at just $2,000 for the 65-inch model I tested. What you get with the UR9 is an improved picture quality compared to the brand’s other models, which are typically priced lower than sets from big names like Samsung, Sony, and LG. I’ve tested countless Hisense entry-level models over the years, including a few that had poor contrast and brightness, putting them more in line with TCL, Roku, and Amazon Fire TV bargain models that cost around $800.
Advertisement
Not the case with this gaming monster, with its unusual-but-welcome 180-Hz refresh rate (330-Hz variable) when you link a high-end computer to the DisplayPort connection on the side. Overall, I was impressed by the picture quality at this price point, even if the UR9 can’t quite compete with the latest (but pricey) Samsung and LG models that use micro RGB tech.
Standard Setup for a Unique Television
Photograph: John Brandon
The all-black, notably thin (only 1.8 inches!) UR9 comes with a stand that’s much easier to assemble than the Sony Bravia 7 Mark II RGB TV. Once in position on my stand, setting up the Google TV operating system was simple, save for dealing with a known bug with the Google Home app’s QR code that required manually entering my Gmail address and password. The UR9 uses Wi-Fi 6E, which is faster than Wi-Fi 6.
Palencia man suspected of links to CARR, Z-Pentest, and NoName057(16), plus helping a Ukrainian hacker flee to Russia
Spanish police have arrested a man they believe is affiliated with at least two pro-Russia hacktivist groups linked to attacks on critical national infrastructure (CNI).
Arrested in March at his home in Palencia, central Spain, the man is suspected of having close ties to CyberArmy of Russia Reborn (CARR) and Z-Pentest, and may have carried out attacks on behalf of NoName057(16).
Advertisement
All three hacktivist groups were named by the UK’s NCSC earlier this year as part of an advisory warning about the dangers these groups pose to Western CNI.
The cyber arm of GCHQ, the UK’s signals intelligence agency, said organizations should not underestimate pro-Russia hacktivist groups, despite them being known largely for relatively low-impact DDoS attacks.
Jonathon Ellison, NCSC director of national resilience, said at the time: “We continue to see Russian-aligned hacktivist groups targeting UK organizations, and although denial-of-service attacks may be technically simple, their impact can be significant.
“By overwhelming important websites and online systems, these attacks can prevent people from accessing the essential services they depend on every day.”
Advertisement
A month earlier, US officials said CARR was working with, or receiving instructions from, Russian military intelligence (GRU).
Policía Nacional first announced the detention of the unidentified man on Monday, although the arrest was made months ago following an FBI tip-off.
In August 2025, the feds alerted Spanish police to the man’s alleged involvement in trying helping a Ukrainian hacker, a member of CARR, flee to Russia via Poland and Belarus
He was said to have provided “logistical and support cover” to facilitate the Ukrainian’s escape.
Advertisement
After the Palencia man’s arrest, officers found evidence suggesting he was in close contact with other members of these pro-Russia hacktivist “terrorist groups.” Police said he assisted in “coordinating actions and providing support” for the different outfits’ activities, including those of NoName057(16).
NoName057(16) has been active since at least 2022, and is known for targeting public and private organizations, NATO countries, and those whose interests do not align with Russia’s.
Police also seized computer equipment from the man’s residence and cryptocurrency storage devices, freezing a wallet suspected of containing proceeds of cybercrime.
The FBI’s Cyber Division said in a statement: “Last December, the FBI announced Operation Red Circus, our ongoing effort to disrupt Russian state-sponsored cyber threats to the United States and our interests abroad. As part of that announcement, the FBI and partners released a joint Cybersecurity Advisory on pro-Russia hacktivist groups conducting opportunistic attacks against critical infrastructure, including the water, agriculture, and energy sectors.
Advertisement
“A mission priority of Operation Red Circus is targeting and arresting individuals for their roles in hacktivist groups such as Cyber Army of Russia Reborn to mitigate planned, malicious cyber-campaigns.
Years of pursuit
Authorities have been hunting pro-Russia hacktivists, particularly CARR members, for years. CARR has been active since at least 2022, when it began with low-level attacks in Ukraine shortly after Russia’s invasion.
The US named Yuliya Vladimirovna Pankratova as CARR’s leader and Denis Olegovich Degtyarenko as its primary hacker in 2024. The pair were sanctioned after CARR was tied to attacks on US and European water facilities earlier that year that specifically targeted human-machine interfaces at water supply, hydroelectric, wastewater, and energy facilities.
CARR also gained access to the SCADA system of a US energy company, which allowed them to control alarms and pumps connected to tanks.
Advertisement
Mandiant previously attributed these attacks to Sandworm, a cyber unit inside Russia’s GRU. However, the sanctions pointed to a hacktivist element and added further color to the relationship between Russia’s military and cybercrime community.
Separately, pro-Russia Ukrainian hacktivist Victoria Eduardovna Dubranova, 33, was extradited to the US late last year after being charged with offenses related to attacks carried out by CARR and NoName057(16).
Dubranova was linked to attacks on water facilities and a Los Angeles meat processing facility in November 2024, which spoiled thousands of pounds of meat and triggered an on-site ammonia leak. ®
By Leda Stawnychko and Mehnaz Rafi
You must be logged in to post a comment Login