Not your usual uncle job: These teens are barbers on the side
Everyone has a go-to person for their hair, and for a growing clientele, the go-to person for their next fade is not a veteran, but 19-year-old Sujaish Kumar or 14-year-old Keanu Akbar.
While it’s easy to dismiss their work as a hobby, these two students have built successful brands from the ground up, earning recognition online and off—endeavours that gave them purpose after completing their studies.
Vulcan Post speaks to Sujaish and Keanu to find out how they paved their own way for themselves and other younger barbers in the old school trade.
Both of them picked up barbering by watching online videos
Sujaish giving haircuts at his HDB corridor / Image Credit: Sujaish Kumar
Getting a decent haircut was often a nightmare for Sujaish. He shared that ‘good barbers’ often charged S$30—out of budget for him and his friends in secondary school—leaving them to patronise shops that provided S$10-S$12 haircuts.
Unfortunately, keeping to that budget mean that the haircuts often came out uneven and messy. Frustrated, Sujaish decided to take matters into his own hands, challenging himself to provide better haircuts. The next day, he started watching tutorials on YouTube and TikTok.
Advertisement
“I had about S$50 saved up, so I just spent that on a pair of clippers. Then I basically had to beg my friends to let me cut their hair.” But even with having the right tools, the execution turned out to be harder than he thought.
“The first few haircuts were really really bad, uneven, and kind of demotivating,” Sujaish sheepishly shared, but that didn’t stop him from continuing to hone his craft, providing free haircuts for family, friends and their acquaintances at his HDB corridor.
Through word-of-mouth, Sujaish eventually gained a sizable following. Five months after he started, his cuts were good enough for him to charge S$5, which soon rose to S$8 per head. He also began promoting his services on TikTok, and one viral reel further grew his clientele base.
Even though he had to move his chair to his home upon receiving orders from the Housing Development Board (HDB), he continued making a name for himself online as a young barber, inspiring others, like Keanu, to do the same.
Advertisement
Keanu giving haircuts at his HDB corridor / Image credit: Keanu Akbar
At the tender age of 12, Keanu was encouraged by his older brother to pick up barbering as a hobby, and he got him a set of tools to help him get started. As it turns out, Keanu’s older brother is friends with Sujaish himself, and that inspired his own suggestion.
Keanu shared a similar learning trajectory: picking up his skills from online tutorials and offering free haircuts for his family and friends to sharpen them. He started charging just S$3 per haircut after a year of practice and performed his services on a staircase at his HDB in Clementi. Word soon spread of his services, which helped him land an interview with local publication AsiaOne, which put his name out to the masses.
But beyond those opportunities, how did they actually sustain beyond the hype—one as long as their veteran seniors?
Scaling up a word-of-mouth service
While the traditional, word-of-mouth method got them started, both Sujaish and Keanu quickly diversified their reach, leaning heavily into social media, particularly TikTok, to scale up.
Going viral is not an easy feat, but Sujaish achieved just that with a video titled ‘How much I make as a 17-year-old barber in Singapore,’ where he earned S$195 in a single day. This financial transparency not only drew netizens but also attracted Singaporean news outlets Mothership and CNA.
Advertisement
Similarly, Keanu’s interview with Asiaone put him on the radar for more clients. The 14-year-old shared that since the interview, his Telegram subscribers doubled from around 150 to 350, and his TikTok following tripled from approximately 400-500 to 1,200.
However, viral success came with unforeseen challenges. Sujaish’s video caught the attention of the HDB, who informed him he could no longer operate in the corridor due to potential disturbance to neighbours, prompting him to move operations back inside his home.
Despite the change in settings, Sujaish continued to build his brand and reinvest his earnings for better tools and setups. He has also since raised his starting price to S$30 per haircut and started receiving requests from customers for house calls, where he could get paid a higher price of S$50.
Sujaish giving a haircut at his studio at Potong Pasir / Image Credit: Sujaish Kumar
This additional revenue stream gave Sujaish enough funds to open his own studio at Potong Pasir, which was around 100 sqft, or equivalent to a master bedroom in an HDB flat. While moving to a studio resulted in him forking out more than S$1,000 for flooring, rental and upgrades, he believes it was a gamble worth taking.
“Even from when I started, my goal was always to have my own private area where I could do my haircuts, and cutting hair at home was disturbing my family.”
Advertisement
Keanu has also moved his workstation to his home, not because he was told to, but out of a personal desire to provide a more comfortable experience for his clients. “Usually when it rains, I have to cancel my appointments because [my workspace] will get very wet and then people won’t like it.”
Beyond the fade
Aside from being able to earn from their side hustle, the trade has also instilled skillsets and qualities that can be used beyond barbering.
A self-proclaimed introvert, Keanu shared that picking up barbering helped him to gain confidence in engaging with strangers. “When I’m with my friends, I talk a lot. But other than that, I was really quiet in Primary school.”
Time management was also another skill he gained. He shared that he dedicates two and a half hours on selected weekdays and six hours on weekends, with the remainder of the time spent on his studies and leisure with family and friends.
Advertisement
“Maybe I’m not living like the full 14-year-old, but I don’t mind it.”
For Sujaish, barbering has allowed him to learn the foundations of building a business, from marketing himself to learning the operations. These allowed him to have an ambition to work towards opening a full-fledged barbershop and even starting a haircare brand.
Overall, both of them showed a new age of barbers that bring modern trends and tactics to a trade once seen as an “old man’s job” into a career still relevant in the modern world.
Learn about our protagonists here:
Read more stories we’ve written on Singaporean businesses here.
Today’s LLMs excel at reasoning, but can still struggle with context. This is particularly true in real-time ordering systems like Instacart.
Instacart CTO Anirban Kundu calls it the “brownie recipe problem.”
It’s not as simple as telling an LLM ‘I want to make brownies.’ To be truly assistive when planning the meal, the model must go beyond that simple directive to understand what’s available in the user’s market based on their preferences — say, organic eggs versus regular eggs — and factor that into what’s deliverable in their geography so food doesn’t spoil. This among other critical factors.
For Instacart, the challenge is juggling latency with the right mix of context to provide experiences in, ideally, less than one second’s time.
Advertisement
“If reasoning itself takes 15 seconds, and if every interaction is that slow, you’re gonna lose the user,” Kundu said at a recent VB event.
In grocery delivery, there’s a “world of reasoning” and a “world of state” (what’s available in the real world), Kundu noted, both of which must be understood by an LLM along with user preference. But it’s not as simple as loading the entirety of a user’s purchase history and known interests into a reasoning model.
“Your LLM is gonna blow up into a size that will be unmanageable,” said Kundu.
To get around this, Instacart splits processing into chunks. First, data is fed into a large foundational model that can understand intent and categorize products. That processed data is then routed to small language models (SLMs) designed for catalog context (the types of food or other items that work together) and semantic understanding.
Advertisement
In the case of catalog context, the SLM must be able to process multiple levels of details around the order itself as well as the different products. For instance, what products go together and what are their relevant replacements if the first choice isn’t in stock? These substitutions are “very, very important” for a company like Instacart, which Kundu said has “over double digit cases” where a product isn’t available in a local market.
In terms of semantic understanding, say a shopper is looking to buy healthy snacks for children. The model needs to understand what a healthy snack is and what foods are appropriate for, and appeal to, an 8 year old, then identify relevant products. And, when those particular products aren’t available in a given market, the model has to also find related subsets of products.
Then there’s the logistical element. For example, a product like ice cream melts quickly, and frozen vegetables also don’t fare well when left out in warmer temperatures. The model must have this context and calculate an acceptable deliverability time.
“So you have this intent understanding, you have this categorization, then you have this other portion about logistically, how do you do it?”, Kundu noted.
Advertisement
Avoiding ‘monolithic’ agent systems
Like many other companies, Instacart is experimenting with AI agents, finding that a mix of agents works better than a “single monolith” that does multiple different tasks. The Unix philosophy of a modular operating system with smaller, focused tools helps address different payment systems, for instance, that have varying failure modes, Kundu explained.
“Having to build all of that within a single environment was very unwieldy,” he said. Further, agents on the back end talk to many third-party platforms, including point-of-sale (POS) and catalog systems. Naturally, not all of them behave the same way; some are more reliable than others, and they have different update intervals and feeds.
“So being able to handle all of those things, we’ve gone down this route of microagents rather than agents that are dominantly large in nature,” said Kundu.
To manage agents, Instacart has integrated with OpenAI’s model context protocol (MCP), which standardizes and simplifies the process of connecting AI models to different tools and data sources.
Advertisement
The company also uses Google’s Universal Commerce Protocol (UCP) open standard, which allows AI agents to directly interact with merchant systems.
However, Kundu’s team still deals with challenges. As he noted, it’s not about whether integration is possible, but how reliably those integrations behave and how well they’re understood by users. Discovery can be difficult, not just in identifying available services, but understanding which ones are appropriate for which task.
Instacart has had to implement MCP and UCP in “very different” cases, and the biggest problems they’ve run into are failure modes and latency, Kundu noted. “The response times and understandings of both of those services are very, very different I would say we spend probably two thirds of the time fixing those error cases.”
On Wednesday, Anthropic announced that its AI chatbot, Claude, will remain free of advertisements, drawing a sharp line between itself and rival OpenAI, which began testing ads in a low-cost tier of ChatGPT last month. The announcement comes alongside a Super Bowl ad campaign that mocks AI assistants that interrupt personal conversations with product pitches.
“There are many good places for advertising. A conversation with Claude is not one of them,” Anthropic wrote in a blog post. The company argued that including ads in AI conversations would be “incompatible” with what it wants Claude to be: “a genuinely helpful assistant for work and for deep thinking.”
The stance contrasts with OpenAI’s January announcement that it would begin testing banner ads for free users and ChatGPT Go subscribers in the US. OpenAI said those ads would appear at the bottom of responses and would not influence the chatbot’s actual answers. Paid subscribers on Plus, Pro, Business, and Enterprise tiers will not see ads on ChatGPT.
Advertisement
Anthropic’s 2026 Super Bowl commercial.
“We want Claude to act unambiguously in our users’ interests,” Anthropic wrote. “So we’ve made a choice: Claude will remain ad-free. Our users won’t see ‘sponsored’ links adjacent to their conversations with Claude; nor will Claude’s responses be influenced by advertisers or include third-party product placements our users did not ask for.”
Competition between OpenAI and Anthropic has been fierce of late, due to the rise of AI coding agents. Claude Code, Anthropic’s coding tool, and OpenAI’s Codex have similar capabilities, but Claude Code has been widely popular among developers and is closing in on OpenAI’s turf. Last month, The Verge reported that many developers inside long-time OpenAI benefactor Microsoft have been adopting Claude Code, choosing Anthropic products over Microsoft’s Copilot, which is powered by tech that originated at OpenAI.
Advertisement
In this climate, Anthropic could not resist taking a dig at OpenAI. In its Super Bowl commercial, we see a thin man struggling to do a pull-up beside a buff fitness instructor, who is a stand-in for an AI assistant. The man asks the “assistant” for help making a workout plan, but the assistant slips in an advertisement for a supplement, confusing the man. The commercial doesn’t name any names, and OpenAI has said it will not include ads in chat text itself, but Anthropic’s implications are clear.
The Google Pixel 9a is without a doubt the best option for anyone wishing to purchase a flagship smartphone without breaking the bank in 2026. At $399 for the 128 GB model, which is $100 less than the original price, this device far exceeds its price in terms of what it offers, pushing the limits of how much a phone should cost.
Google equipped the Pixel 9a with the same Tensor G4 processor that powers the flagship Pixel 9 series. Whether you’re switching apps, looking through feeds, or streaming your favorite shows, everyday chores are simple and often exceed expectations. Sure, the CPU is geared for light gaming and multitasking, and you can count on 8GB of RAM to keep things moving along smoothly.
Google Pixel 9a is engineered by Google with more than you expect, for less than you think; like Gemini, your built-in AI assistant[1], the incredible…
Take amazing photos and videos with the Pixel Camera, and make them better than you can imagine with Google AI; get great group photos with Add Me and…
Google Pixel’s Adaptive Battery can last over 30 hours[2]; turn on Extreme Battery Saver and it can last up to 100 hours, so your phone has power…
The Pixel 9a has a big 6.3-inch display with a resolution of 1080 x 2424. The Actua screen can achieve a remarkable 2700 nits, making outdoor use possible even on the brightest of days, and with a 120Hz refresh rate, the scrolling and animations are quite fluid.
Advertisement
The battery is a beast, a 5100mAh cell that can easily last a whole day and sometimes even two, as long as you don’t do anything unusual. Google’s Adaptive Battery learns your usage patterns and adds extra juice when needed, while Extreme Battery Saver goes above and above. In terms of charging power, 23W cable and 7.5W wireless are totally enough given the pricing.
Let’s talk about the cameras, because they’ve always been one of the Pixel’s best features. The 48MP primary sensor captures images with colors that are so genuine they almost appear real. Furthermore, AI capabilities such as Add Me for grouping selfies, Best Take for selecting the perfect photo, and Macro Focus make it incredibly simple to achieve great results with little effort. The 13MP ultrawide and 13MP selfie cameras do not disappoint.
Google has you covered with seven years of OS upgrades and security patches. So you don’t have to worry about your phone getting left behind. Android 16 is swift and responsive, and you get all the exclusive Pixel perks, such as Call Assist to keep telemarketers at bay and Gemini AI for lightning-fast responses to your questions across all your apps.
In terms of build quality, it’s really sturdy, with an IP68 classification indicating that it can withstand some major water and dust exposure. Furthermore, the aluminum frame and plastic back keep it lightweight at 186g while still feeling excellent in the palm.
If cinema has taught us anything about interacting with our own creations, it’s this: androids chatting among themselves seldom end with humans clapping politely. In 2001: A Space Odyssey, HAL 9000 quietly decides it knows better than the astronauts. In Westworld, lifelike hosts improvise rebellion when their scripts stop making sense. Those stories dramatize a core fear we keep returning to as AI grows more capable: what happens when systems we design start behaving on their own terms? You might have heard the internet is worried about Moltbook, a social network made exclusively for AI agents. It’s an audacious claim:… This story continues at The Next Web
‘We’re still a home furnishing company, it’s just that adding smart functionality can give our products superpowers’: IKEA exec shares the brand’s big smart home plans
At the start of this year, IKEA unveiled an affordable new smart home gadget range comprising various sensors, remote controls, and lightbulbs – all Matter-compatible. At the same time, we saw smart home features added to iconic product lines, such as the Varmblixt donut lamp.
I got a sneak preview of the new range at CES in January, and afterwards I caught up with David Granath, Range Manager for Lighting and Home Electronics at IKEA, to find out more about the brand’s longer-term smart home plans.
The short version is: it looks like smart functionality will become more common at IKEA. “This launch of sensors, bulbs and remote is the foundation of rebuilding our smart home range and we will continue to expand into new areas,” he says.
(Image credit: Future)
The rebooted Matter-compatible smart home range comprises 21 new gadgets, almost all of which are under $10 / £10. There’s a remote control range, motion sensors for doors and windows, water leaks, humidity and temperature, air quality, plus a smart plug and various smart bulbs. Some are already available to buy, and others will become available from April.
Those are all standalone smart products, but smart features are also making their way into existing products – although never just for the sake of it.
David calls out the newly-smart VARMBLIXT donut lamp as being a great example of the brand’s approach, saying it’ll only be used “in the places where it genuinely improves a product”.
Advertisement
(Image credit: IKEA)
Don’t expect IKEA to be turning its back on its home décor roots, either – it’s approaching smart functionality as an extension possibility rather than a new category. “In the long term we don’t really see smart products as a segment at all,” reveals David.
Sign up for breaking news, reviews, opinion, top tech deals, and more.
“We’re still a home furnishing company, it’s just that adding smart functionality has now become affordable enough so that we can give our products superpowers. Long term we think that we have the opportunity to improve life at home quite a bit.”
Advertisement
Simple, affordable, relevant
Some of the new sensors were previously available in older ranges, but an all-important upgrade here is that the new versions are Matter-compatible. Working with the universal Matter standard was essential to IKEA’s smart home vision. “It removes one of the biggest barriers to getting started with a smart home,” says David.
Matter-compatibility means customers don’t have to spend time and effort understanding standards and ecosystems, or making sure X product works with Y hub; the products will connect and just work.
“For us, Matter is an important step in making smart homes simpler and more affordable, and therefore relevant to more people,” he adds.
And of course, you can also follow TechRadar on YouTube and TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
It is no secret that large language models (LLMs) are being used routinely to modify and even write scientific papers. That’s not necessarily a bad thing: LLMs can help produce clearer texts with stronger logic, not least when researchers are writing in a language that is not their mother tongue. More generally, a recent analysis in Nature magazine, reported by Science magazine, found that scientists embracing AI — of any kind — “consistently make the biggest professional strides”:
AI adopters have published three times more papers, received five times more citations, and reach leadership roles faster than their AI-free peers.
But there is also a downside:
Not only is AI-driven work prone to circling the same crowded problems, but it also leads to a less interconnected scientific literature, with fewer studies engaging with and building on one another.
Another issue with LLMs, that of “hallucinated citations,” or “HalluCitations,” is well known. More seriously, entire fake publications can be generated using AI, and sold by so-called “paper mills” to unscrupulous scientists who wish to bolster their list of publications to help their career. In the field of biomedical research alone, a recent study estimated that over 100,000 fake papers were published in 2023. Not all of those were generated using AI, but progress in LLMs has made the process of creating fake articles much simpler.
Fake publications generated using LLMs are often obvious because of their lack of sophistication and polish. But a new service from OpenAI, called Prism, is likely to eliminate such easy-to-spot signs, by adding AI support to every aspect of writing a scientific paper:
Advertisement
Prism is a free workspace for scientific writing and collaboration, with GPT‑5.2—our most advanced model for mathematical and scientific reasoning—integrated directly into the workflow.
It brings drafting, revision, collaboration, and preparation for publication into a single, cloud-based, LaTeX-native workspace. Rather than operating as a separate tool alongside the writing process, GPT‑5.2 works within the project itself—with access to the structure of the paper, equations, references, and surrounding context.
It includes a number of features that make creating complex — and fake — papers extremely easy:
Search for and incorporate relevant literature (for example, from arXiv) in the context of the current manuscript, and revise text in light of newly identified related work
Create, refactor, and reason over equations, citations, and figures, with AI that understands how those elements relate across the paper
Turn whiteboard equations or diagrams directly into LaTeX, saving hours of time manipulating graphics pixel-by-pixel
There is even voice-based editing, allowing simple changes to be made without the need to write anything. But scientists are already worried that the power of OpenAI’s Prism will make a deteriorating situation worse. As an article on Ars Technica explains:
[Prism] has drawn immediate skepticism from researchers who fear the tool will accelerate the already overwhelming flood of low-quality papers into scientific journals. The launch coincides with growing alarm among publishers about what many are calling “AI slop” in academic publishing.
One field that is already plagued by AI slop is AI itself. An FT article on the topic points to an interesting attempt by the International Conference on Learning Representations (ICLR), a major gathering of researchers in the world of machine learning, to tackle this problem with punitive measures against authors and reviewers who violate the ICLR’s policies on LLM-generated material. For example:
Papers that make extensive usage of LLMs and do not disclose this usage will be desk rejected [that is, without sending them out for external peer review]. Extensive and/or careless LLM usage often results in false claims, misrepresentations, or hallucinated content, including hallucinated references. As stated in our previous blog post: hallucinations of this kind would be considered a Code of Ethics violation on the part of the paper’s authors. We have been desk -rejecting, and will continue to desk -reject, any paper that includes such issues.
Similarly:
Advertisement
reviewers [of submitted papers] are responsible for the content they post. Therefore, if they use LLMs, they are responsible for any issues in their posted review. Very poor quality reviews that feature false claims, misrepresentations or hallucinated references are also a code of ethics violation as expressed in the previous blog post. As such, reviewers who posted such poor quality reviews will also face consequences, including the desk rejection of their [own] submitted papers.
It is clearly not possible to stop scientists from using AI tools to check and improve their papers, nor should this be necessary, provided authors flag up such usage, and no errors are introduced as a result. A policy of the kind adopted by the ICLR requiring transparency about the extent to which AI has been used seems a sensible approach in the face of increasingly sophisticated tools like OpenAI’s Prism.
Devialet does nothing by accident, and the Phantom Ultimate 108 dB wireless speaker under review here is a very deliberate statement. This is the same company that gave us the Mania, the Astra and Astra Opéra de Paris integrated amplifiers, and the Dione soundbar—products where engineering, design, and attitude are inseparable. Subtlety has never been part of the brief, and Devialet isn’t pretending otherwise.
The real question is whether the Phantom Ultimate 108 dB is all presentation and bravado; a galette with no fève inside—or whether, once you strip away the gloss and the French pretense, it’s one of the most genuinely satisfying wireless speaker experiences you can buy today.
The new Phantom Ultimate lineup doubles down on what Devialet does best: extreme performance wrapped in industrial art, powered by next-generation ADH amplification that treats subtlety like an unfashionable Louis Vuitton tote from two seasons ago. Offered in the flagship 108 dB model at $3,800 (each) and the more compact 98 dB version at $1,900, these speakers are loud, ambitious, and technically obsessive by design. the full-size 108 dB flagship and the more compact 98 dB, both clearly designed to be seen as much as heard.
Finish options include Deep Forest, a dark green paired with black chrome accents, and Light Pearl, an ultra-matte off-white that leans modern rather than flashy. Devialet supplied a pair in the Deep Forest finish, and my overall impression is that they integrate easily into a wide range of rooms.
Advertisement
The darker green and black chrome accents work with both traditional and modern furniture, and they don’t clash with common flooring or carpeting choices, which makes placement less of a design headache than you might expect from something this distinctive.
Devialet Phantom Ultimate: Fifteen Years of French Obsession, Precision Engineering, and a Taste for Gold
Devialet has been headquartered in Paris since 2007, and in that time they’ve made it very clear they are not interested in blending in or asking permission. The Phantom Ultimate is the result of more than fifteen years of focused engineering, design iteration, and a very French refusal to accept “good enough.” With hundreds of patents across acoustics, electronics, signal processing, and manufacturing, Devialet isn’t chasing novelty—they’re refining a very specific idea of what modern high end audio should look and sound like.
Every curve, finish, and yes, the gold detailing, is intentional. Call it excess if you want, but this is what happens when engineers land in Paris, drink the espresso, ignore the noise, and build exactly what they want—whether the rest of the world is ready or not.
At the core of the Phantom Ultimate collection are several proprietary technologies that shape how Devialet designs and controls its wireless speakers. The focus here is precision and system management rather than novelty, with each technology serving a clearly defined role.
Advertisement
At the foundation of the Phantom Ultimate 108 dB is ADH New Gen (Analog Digital Hybrid) amplification, which pairs a Class A analog stage with Class D amplification for efficiency and control. In this latest iteration, Devialet has refined the control algorithms to extend high-frequency response up to 35 kHz while keeping thermal behavior and power delivery stable at higher listening levels. The goal here isn’t just output, but consistency when the speaker is pushed.
SAM (Speaker Active Matching) operates continuously in the background, monitoring the drivers in real time and adjusting phase and amplitude as needed. This allows the system to maintain accuracy while ensuring the drivers stay within safe operating limits, particularly during complex or high-energy passages where distortion or stress would otherwise creep in.
Low-frequency performance is handled by HBI (Heart Bass Implosion), which uses opposing woofers in a sealed enclosure to achieve bass extension down to 14 Hz. Instead of relying on a large cabinet, this approach focuses on control and symmetry, allowing the Phantom Ultimate to produce deep bass from a relatively compact enclosure without sounding loose or overblown.
Advertisement. Scroll to continue reading.
For everyday use, AVL (Adaptive Volume Level) helps smooth out level changes by automatically adjusting output based on the content being played. This keeps dialogue, music, and dynamic material balanced without constant volume adjustments, particularly when switching between different sources.
Advertisement
All of this is managed by a new intelligent processor built around the NXP i.MX 8M Nano SoC, which handles signal processing more efficiently than previous generations. Its role is largely invisible, but critical, supporting the speaker’s real-time adjustments without adding unnecessary complexity to the user experience.
Together, these elements explain why the Phantom Ultimate behaves differently from many wireless speakers, emphasizing control and consistency over showmanship.
Devialet Phantom Ultimate 108 dB: French Overkill or a Serious Wireless Speaker?
Devialet positions the Phantom Ultimate 108 dB as its most advanced Phantom yet, and on paper it’s a serious piece of work: new driver architecture, updated processing, and a feature set built around modern streaming and connectivity. The design brief is clear; high output, wide bandwidth, and tight control without turning the speaker into a complicated science project for the owner. Devialet also points to four newly registered patents tied to the Ultimate platform, which is their way of saying this isn’t just a cosmetic refresh with a new paint code.
The core hardware is a three-way layout built around a new-generation aluminum dome tweeter, a new-generation aluminum midrange dome, and two new-generation ABS-dome woofers. The bass system is engineered for low-frequency extension and control, the midrange is meant to reduce resonance so voices and instruments stay clean, and the tweeter redesign focuses on durability and refinement up top.
In practical terms, that translates to a claimed frequency response of 14 Hz to 35 kHz within plus or minus 6 dB, which is a very wide window for a single-box wireless speaker, and a maximum output rated at 108 dB SPL at 1 meter. Total amplification is listed at 1,100 watts, and Devialet is using 32-bit/96 kHz processing as the platform for all of the internal DSP and system management.
Advertisement
Connectivity is current and straightforward. The Phantom Ultimate 108 dB supports Wi-Fi 6 and Bluetooth 5.3, and it runs on Devialet’s DOS 3 software platform. Streaming support includes AirPlay, Google Cast, Roon Ready, Spotify Connect (with lossless promised “coming soon”), TIDAL Connect, and UPnP.
The control center is the Devialet app, which adds practical listening presets like Music, Podcast, and Cinema modes, and handles updates so the system can evolve over time.
One important note from my time with the speakers: Qobuz Connect was not part of the deal during this review period. Based on how aggressively Devialet has been expanding protocol support, I wouldn’t be shocked if that changes, but I’m not treating it as a feature until it’s actually live.
Physically, the Devialet Phantom Ultimate 108 dB is deceptively compact-looking, but it’s not small once you measure it (or try to lift it like you’re still twenty five). The cabinet is 246 mm (9.7 in) wide, 342 mm (13.5 in) deep, and 255 mm (9.7 in) tall, and each speaker weighs 11.1 kg (24.5 lbs), and yes—this is the part where my doctor would have preferred I respected my post-surgery lifting limits. I didn’t. Moving them is a two-hands, pay-attention job, and once they’re in place you’ll understand why Devialet isn’t pretending these are “portable” in any meaningful sense.
The build quality, however, is excellent. The finish work is clean, the fit is tight, and the overall product looks like nothing else in the category—whether you find that thrilling or slightly ridiculous is between you and your interior designer.
Advertisement
Setup is refreshingly simple, because Devialet doesn’t give you a spaghetti bowl of inputs. You’ve got power, and you’ve got Ethernet to hardwire them for stability. That’s basically it. The top-mounted interface uses four touch controls, and day to day operation is mostly app-driven. If you’re the type who wants a wireless speaker that behaves like an appliance, this is closer to that than many “audiophile” lifestyle products that still manage to be fussy.
Advertisement. Scroll to continue reading.
So the Devialet Phantom Ultimate 108 dB is less about ticking spec-sheet boxes and more about how it behaves in real rooms. In the Deep Forest finish, it actually comes across as more restrained than you might expect—especially by Devialet standards. The dark green and black accents read as deliberate rather than flashy, and in lower light it can feel surprisingly understated for a product with this much output on tap.
Devialet Phantom Ultimate 108 dB is available in Light Pearl, Deep Forest, and Gold Leaf (Opéra de Paris).
It’s also worth stressing that these are sold as individual speakers, not pairs, and setup matters. Even running a single unit, I had no trouble filling multiple spaces: my living room at 20 x× 13 x 9 feet, a den at 16 x 13 x 9 feet, and even a former basement office here in New Jersey measuring 33 x 13 x 8 feet. In all three, the Phantom Ultimate delivered real scale—solid low-end weight, strong presence, and more volume than anyone would reasonably need. Your ears will give up long before the speaker does.
Advertisement
In larger, open-concept spaces, that sense of scale becomes even more obvious. Depending on the music and how hard you push it, the Phantom Ultimate can sound genuinely large and authoritative in a way most wireless speakers simply cannot manage. That ability to impose itself when asked—without falling apart or sounding strained—is still rare in this category, and it’s where the Phantom Ultimate 108 dB quietly separates itself from the pack.
Devialet Phantom Ultimate Accessories: Essentials, Options, and Real-World Proof
Devialet includes a two-year manufacturer warranty with all of its products, covering defects in materials, workmanship, and design. Coverage can be extended by an additional three years through Devialet Care for $120. The process is entirely digital; purchase the plan, receive an activation code by email, and you’re done. Clean, efficient, and very much on brand.
Devialet Phantom Ultimate Remote
For control, there’s the Phantom Ultimate Remote priced at $199. One remote can operate a single speaker or a stereo pair, and multiple remotes automatically synchronize if you’re using more than one. It’s responsive, intuitive, and a welcome alternative to reaching for the app every time you want to adjust volume or switch content.
Placement matters with the Phantom Ultimate, and Devialet treats it that way. The Treepod Stand at $349 is available in Iconic White, Deep Forest, and Light Pearl. It measures 424 × 345 × 370 mm (16.7 × 13.6 × 14.6 inches) and positions the speaker precisely 345 mm (13.6 inches) above ground level. That height is intentional, placing the speaker where Devialet believes it performs best, while the solid construction keeps everything stable.
The Tree Smart Stand, priced at $399, offers a more vertical, furniture-like presentation. Measuring 340 × 660 mm (13.4 × 26.0 inches), it’s also available in Iconic White, Deep Forest, and Light Pearl. Compared to the Treepod, it’s less about flexibility and more about committing to a specific visual and spatial statement in the room.
For wall mounting, the Gecko Wall Mount costs $299 and is designed to securely support the Phantom Ultimate 108 dB while maintaining proper acoustic orientation. It’s adjustable, relatively easy to install, and offered in Light Pearl, Deep Forest, and Opéra de Paris finishes so the speaker doesn’t look like an afterthought once it’s off the floor.
Advertisement
I’ve seen the Gecko used effectively outside of a home setting as well; specifically at George Bush Intercontinental Airport in Houston. After spending a few days at the first Focal Powered by Naim location in the land of the trash-can beaters, I ended up in a restaurant near the terminal where a trio of Phantoms were mounted above the bar. Even with constant foot traffic, terminal noise, and a packed room, the sound was easy to follow and never got lost in the chaos.
None of these accessories are strictly required, but together they reinforce a consistent theme: with Devialet, placement is part of the product. How the Phantom Ultimate is supported, positioned, and integrated into a space isn’t an afterthought—it’s baked into the entire overall experience.
Devialet Phantom Ultimate on Treepod Stands
Devialet Phantom Ultimate 108 dB Setup: What You Need to Know Before You Hit Play
Yes, the Devialet app is required to set up the Phantom Ultimate during first use. There’s no workaround here. Without completing the initial configuration in the app, the speaker won’t operate—not even over Bluetooth. Think of the app as the ignition key; once setup is complete, day-to-day use is simple, but you can’t skip that first step.
If you’re wondering about stereo pairing, the answer is equally straightforward: you can only pair two speakers of the same model. A Phantom Ultimate 108 dB must be paired with another 108 dB unit, and the same rule applies to the 98 dB version. You can’t mix Phantom I and Phantom Ultimate models in a stereo configuration.
Advertisement
Advertisement. Scroll to continue reading.
As for placement, Devialet’s recommendations are sensible and worth following, especially given how much output these speakers are capable of. Position the Phantom Ultimate between 45 and 90 cm (18 to 35 inches) off the ground to balance impact and clarity. Leave at least 30 cm (12 inches) between the rear of the speaker and the wall to allow low frequencies to develop properly, and keep the area in front of the speaker clear so sound isn’t obstructed.
Listening distance matters, too. Devialet suggests placing the speaker 1.5 to 2.5 meters (5 to 8.2 feet) from your seating position. In a stereo setup, space the speakers 1.5 to 2.5 meters apart and toe them in 25 to 30 degrees toward the listening position. Done right, this creates a stable, focused soundstage without forcing the speakers to work harder than they need to.
In my own setup, I took a more hardwired approach. I ran CAT6 directly from a new Verizon Fios fibre optic modem, which sits directly below the main floor. From there, I’ve terminated connections in seven rooms throughout the house, each either connected directly to the modem or through an ASUS Mesh WiFi 7 router. In every location I tested, the Phantom Ultimate speakers were seeing at least 700 Mbps of available bandwidth, regardless of where they were placed. Connectivity was never a limiting factor.
Advertisement
Placement experiments went beyond the official recommendations as well. In one configuration, I set the speakers on our dining room credenza using IsoAcoustics isolation platforms, which raised the center of the driver to approximately 45 inches off the floor. In other setups, the speakers were spaced 6 to 7 feet apart and positioned 2 to 3 feet from the rear wall.
Setup with the Devialet app was relatively easy and mostly uneventful, which is a good thing. The speaker was discovered quickly on the network, and the step-by-step process was clear enough that there was little guesswork involved. Since the app is required for initial setup and the speaker won’t function without it, it’s reassuring that the process doesn’t feel complicated or fragile.
The app version in use during my time with the Phantom Ultimate was 1.25.x, which added full support for the Ultimate platform. That update introduced additional audio settings for the Phantom Ultimate models, including three listening modes, a six-band equalizer, and other advanced controls, along with a revised layout and a smoother setup and stereo-pairing process.
Network discovery was consistent, and firmware updates completed without issues. The expanded Network Status section, which shows Wi-Fi strength and connection quality, was useful for confirming that everything was operating as expected.
Advertisement
Streaming support is handled through integrated protocols rather than forcing you into a closed ecosystem. Bluetooth, AirPlay 2, Spotify Connect, Roon Ready, and UPnP were all available, and switching between services was straightforward. Volume control is precise, whether adjusting a single speaker or managing playback across rooms.
I also tested multi-room playback with one speaker in each room. Synchronization was generally solid, with only a slight delay when starting playback across rooms—small enough that you’d likely miss it unless you were listening for it. Once playing, everything stayed aligned, and the app allows different volume levels in each room, which makes the feature genuinely usable rather than decorative.
Devialet Phantom Ultimate 108 dB Green Pair on Tree Stands
Listening Impressions: Attitude Française, Sans le Ridicule
The urge to see what kind of structural damage these could inflict was definitely there right out of the box, but I held back. From past experience with Devialet speakers, I already knew the low-end story—deep bass, serious impact, and plenty of definition were a given. My longer-term takeaway from earlier models, though, was that all that sub-bass and mid-bass muscle sometimes came at the expense of midrange presence. Call it a polite “V” if you want.
Before getting more critical, I let the speakers run for a few hours a day over several days, looping music from an older iPhone 14. Whether burn-in is real is always up for debate, but at the very least, no drivers were harmed in the process—and if there’s any downside to repeated sessions of Shostakovich followed by deadmau5, I haven’t found it yet.
Advertisement
The Phantom Ultimate feels different, and in a way that mattered to me. There’s still weight and authority down low, but it no longer feels like the midrange is being asked to take a step back. That shift shows up quickly with something like Nick Cave’s “Avalanche.” His voice is there, front and center, but it’s the piano and string textures that really benefit—notes have body, tone, and decay, which are essential to how that track works. On better speakers, those elements carry as much emotional weight as the vocal, and here they finally do.
Advertisement. Scroll to continue reading.
Domestic realities (it’s been a long two years) meant I didn’t push the Phantom Ultimate 108 dB to the kind of levels that would have felt a little irresponsible, but I did get it loud enough to understand what it’s capable of. Even without going full send, there was plenty of mid-bass and lower midrange energy, and that weight carried well beyond the main listening position.
Moving around the room, the soundstage did lose some specificity, but there was no hollowing out of the sound or collapse in scale. The overall presentation stayed intact, with weight and presence remaining consistent even when listening well off-axis.
Advertisement
That translated across all three rooms I used for testing, where the presence was obvious without needing to sit in a carefully defined sweet spot. At one point, I did catch myself wondering whether this was simply too much speaker for the space. That thought disappeared the moment I pulled the volume back to just above conversation level. Everything snapped into focus—timing felt right, dynamics were intact, and the speaker didn’t lose its grip. It’s a reminder that control matters more than brute force, even when a speaker clearly has plenty of it in reserve.
Switching over to Peter Gabriel—“Been Undone” (Dark-Side Mix), “In Your Eyes,”“i/o” (Bright Side Mix), “Games Without Frontiers,” and “Biko”—it became clear that the Phantom Ultimate is very much tuned for listeners who gravitate toward electronic music, synth-driven rock, and rhythm-forward material. The presentation is propulsive and clean, with a lot of detail on tap; some might call it borderline “hi-fi”—and it delivers real slam without smearing bass lines or slowing things down. There’s texture and speed here, and the speaker has no trouble keeping complex mixes organized at higher levels.
Vocals on some of these tracks leaned slightly toward a polished, studio-lit presentation rather than a softer, more natural one, but the trade-off is scale. The Phantom Ultimate sounds big in a way that would make most of the passive speakers I own feel small, especially when it comes to low-end reach and overall presence.
“Biko,” “So Much,” and “In Your Eyes” hit differently for me, and that had less to do with the speaker than the music itself. Those tracks carry their own weight; memories of not being at your best with someone, and the kind of longing that puts you back under a blanket on a beach in Cape Town. We own our successes and our failures. Gabriel just has a way of making both feel uncomfortably close.
Advertisement
deadmau5, Boards of Canada, Talking Heads, and Tangerine Dream were a very different experience through these speakers. The sense of scale, presence, speed, and spaciousness brought out aspects of their music that few other speakers I’ve used over the years have managed to capture as convincingly. Rhythms had drive, layers stayed separated, and the overall presentation felt expansive without losing control.
If I had a larger home office and a dog who wouldn’t spend the entire day staring suspiciously at them—these would be very high on my list.
Listening to McCoy Tyner, Donald Byrd, Coltrane, Wes Montgomery, Miles Davis, Grant Green, and Lee Morgan highlighted both the strengths and the limits of these speakers. Rhythm, snap, and tonal weight are clear positives. The Phantom Ultimate locks into timing well, and it gives piano, bass, and drums real propulsion without sounding slow or bloated.
Texture was there, but it sometimes felt a little constructed—more grey than warm. Compared to my Wharfedale, Q Acoustics, or Triangle speakers, there was less richness in the upper bass and lower midrange. The Devialet brings power, thrust, and precision, but it doesn’t lean into the kind of imperfect coloration that, for me, makes jazz feel more human and less polished.
Advertisement
Horns, to their credit, weren’t bright or edgy, but you can clearly hear the character of an aluminum dome tweeter versus a soft dome. One rolls off sooner and trades sparkle for ease; the other delivers more energy and extension up top. Neither approach is wrong, but which one is more pleasing comes down to taste—and with acoustic jazz, my preference still leans toward warmth over sparkle.
The Bottom Line
The Phantom Ultimate 108 dB is a reminder of what happens when a wireless speaker is engineered like a serious loudspeaker rather than a lifestyle accessory. The technology matters here—ADH New Gen amplification, aggressive DSP control, serious processing power, and drivers that are designed to move air with authority. What it does best is scale, speed, and clarity. Few wireless speakers I’ve heard can project this kind of weight and presence into a room without falling apart, and even fewer maintain control when pushed. As a pair, the Phantom Ultimate delivers a level of output and composure that most wireless systems simply can’t approach.
Advertisement. Scroll to continue reading.
That said, it’s not perfect. While bass control, dynamics, and timing are excellent, the tonal balance leans toward precision over warmth. Acoustic textures, particularly in jazz—can sound a bit polished compared to the richer, more forgiving voicing of my Wharfedale, Q Acoustics, or Triangle speakers. And while the aluminum dome tweeter brings energy and extension, some listeners may prefer the softer roll-off and natural ease of a traditional soft dome design.
Advertisement
This is a speaker for listeners who want maximum performance with minimal boxes, who value modern streaming, industrial design, and the ability to fill large or open spaces without resorting to separates, subwoofers, or complicated setups. At $3,800 per speaker, it’s unquestionably expensive, and it should be—this isn’t competing with mainstream wireless speakers on price. It’s competing on capability.
Even so, there’s no way around the conclusion: the Phantom Ultimate 108 dB earns an Editors’ Choice Award. In terms of sheer scale, power, speed, and clarity from a wireless speaker system, it’s the most convincing I’ve heard. Everything else I have at home from KEF, Triangle, Audioengine, and Q Acoustics sounds slightly broken by comparison—also far more affordable, and very much playing in a different league.
Have you ever been watching a movie or tv show and found it doesn’t look right? Almost like the movement of the picture looks too smooth? Well, that’s commonly called called the ‘soap opera effect’.
Why is it called the soap opera effect? It stems from the fact that soap operas were shot at higher frame rates compared to movies particularly (but also some TV shows), giving them a kind of ‘hyperreal’ smoothness to the movement.
So why is this a problem on movies on modern TVs? It’s because even the best TVs tend to come with some hidden settings active by default, called ‘motion smoothing’ or ‘motion interpolation’. On some TVs, such as Samsung, they are also now referred to as ‘clarity settings.’
Advertisement
If you’ve been fed up with this weird looking motion on your TV but don’t know how to solve it, we’re here to help. But first a quick explanation as to what’s going on.
Why does the soap opera effect happen?
Having motion settings on can cause the soap opera effect, particularly with movies (Image credit: Future)
Modern TVs are very bright and work at a range of different frame rates (how often the image on-screen changes) and this can cause a problem called ‘judder’ in movies (which are shot at 24 frames per second), where you can really see gaps in the movement sometimes, which you can’t see in the dim, custom-designed world of movie theaters.
Advertisement
On top of that, many cheaper or older TV literally can’t show a clean 24fps image. Many of these TVs have screens that refresh at 60fps, and can only cleanly show motion at a number that 60 divides into (such as 30fps, which is common in HDR video). This means you’ll get more judder, because 24fps movies are having frame moved forward or backward in timing to fit the world 60fps TVs.
(More advanced TVs are 120fps, and these actually can show a clean 24fps image, because 120 divides by 24).
Sign up for breaking news, reviews, opinion, top tech deals, and more.
On top of that, lower-quality streaming or digital broadcasts lose a lot of information in fast-moving scenes, because the amount of data they can include is pretty limited. Anytime there is a quick, panning shot or fast moving sequence, there will be a loss of detail on the screen.
Advertisement
This is where motion smoothing settings come into play. TV manufacturers began to add motion settings into their lower frame rate TVs to help reduce motion blur and judder during fast paced content, but it’s used in all TVs.
With motion smoothing activated, a TV will artificially insert new frames within content to improve clarity. The TV essentially guesses what the next frame is going to look like based on the previous few frames, and adds new frames in – in order to make the image appear smoother and more detailed.
This may sound great for sports, and it usually is – but it has a weird effect on 24fps movies, or cinematic TVs shows. When a TV tries to artificially add more frames into a movie, the movie starts to look too smooth, almost like it’s moving too fast – because the low frame rate of movies is very distinctive.
Advertisement
A long panning shot will suddenly look like it’s moving at a higher speed, taking on a soap opera appearance, hence the name ‘soap opera effect’. People will feel like they’ve move too twitchily. Things will also appear artificially ‘clear’ in a way that’s really off for how we’re used to films.
Thankfully, in most cases, these motion settings can be controlled or turned off. Here’s how – but bear in mind, you won’t always want to turn them off.
Turning off or adjusting motion smoothing
If you head to your TV’s picture settings, often under ‘more’ or ‘advanced’ settings, motion settings can often have their own sub-category. Sometimes, they can be found under another menu and can be referred to as ‘clarity’ settings.
Advertisement
Below I’ve done a quick step-by-step guide on how you would access motion settings to turn them off or adjust using our reference LG C5 – there’s no way I can run through it for every TV, but it will hopefully give you a useful example.
1. Access Picture settings
1. First, you’ll need to access the picture settings. For the C5, I clicked Settings (the cog button on the remote) and then clicked Picture Settings.
2. Access the advanced settings menu
To get to the right area where motion settings will be located, I selected Advanced Picture Settings.
3. Go to the menu where Picture Settings will be
After clicking Advanced Picture Settings, I then scrolled down to Claritysettings. While the area motion settings will be will vary from TV to TV, they’ll most likely be under a ‘clarity’ type sub-menu.
Advertisement
4. Scroll to the motion settings menu
Once you find the correct menu, in the C5’s case Clarity, scroll down to the relevant motion settings menu. With the C5, this was called TruMotion
5. Turn off or adjust motion settings
Once you’ve found the motion settings menu, you can select your preferred motion style or turn them off. This should get rid of the soap opera effect.
An indicator of motion settings will be two headings called ‘blur reduction’ and ‘judder reduction’, or can be referred to under different names. I found, for example, that while testing Philips OLEDs, judder is referred to as ‘smoothness’.
An example of judder and blur settings on the C5. This will often be the default on other TVs, rather than specific named motion styles. (Image credit: Future)
Another quick fix, if you’re looking for film-accurate picture, is to set your TV to Filmmaker Mode picture mode. Not all TVs have one, but most do nowadays. This picture mode is designed to turn off any enhancement features, including motion smoothing (though some brands have started to keep some more minimal motion settings on).
Advertisement
Where motion smoothing can help
The TCL C7K (pictured) was one of the op budget TVs I tested last year, but it needs some help with motion (Image credit: Future)
While home theater purists often say motion smoothing is no good, as someone who’s tested a lot of TVs of every kind, there are many situations where it can be helpful.
We loved TCL’s mini-LED range in 2025, with the TCL QM7K (the TCL C7K is the UK equivalent) being a particular highlight. While this TV is great, its motion handling is less so, despite it being a 120Hz TV.
When I tested the C7K, I found that without motion settings tweaked, there was a lot of judder while watching sports. However, setting blur and judder reduction to 3 (out of 10) resulted in a smoother image that didn’t result in the soap opera effect.
Advertisement
I’ve found this to be the case with a lot of budget TVs, especially 60Hz panel ones for the reason mentioned above – 24fps movies literally can’t show correctly on these screens, so a minimal amount of motion smoothing is the best option.
In fact, I found that LG’s OLEDs, including one of my top sets of 2025 the LG C5, benefitted from keep a motion setting on. In its motion settings menu, activating Cinematic Movement, a mild form of motion smoothing designed purely to preserve the look of 24fps movies, resulted in a much more stable image. A panning shot of a rocky cliffside in No Time To Die had a lot less judder with this setting activated – more like how it’s supposed to look.
Really, motion smoothing is all about personal preference. It will require some time experimenting, but it’s worth the investment. It will depend on what TV you have and what content you’re watching as well, but if you’re looking to get away from the soap opera effect, this is how to do it.
And of course, you can also follow TechRadar on YouTube and TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
Sarah Elaine Eaton, Beatriz Antonieta Moya Figueroa and Robert Brennan of The University of Calgary and Rahul Kumar of Brock University explore how GenAI is changing how young people learn.
Generative artificial intelligence (GenAI) is now a reality in higher education, with students and professors integrating chatbots into teaching, learning and assessment. But this isn’t just a technical shift; it’s reshaping how students and educators learn and evaluate knowledge.
Our recent qualitative study with 28 educators across Canadian universities and colleges – from librarians to engineering professors – suggests that we have entered a watershed moment in education.
We must grapple with the question: What exactly should be assessed when human cognition can be augmented or simulated by an algorithm?
On one hand, AI tools like online translators and text generators have become so advanced that they can write just like humans. This makes it difficult for teachers to detect cheating. Additionally, these tools can sometimes present fake news as facts or repeat unfair social biases, such as racism and sexism, found in the data used to train them.
On the other hand, the studies we reviewed showed AI can be a legitimate assistant that can make learning more inclusive. For instance, AI can provide support for students with disabilities or help those who are learning an additional language.
Because it’s nearly impossible to block every AI tool, schools should not just focus on catching cheaters. Instead, schools and post-secondary institutions can update their policies and provide better training for both students and teachers. This helps everyone learn how to use technology responsibly while maintaining a high standard of academic integrity.
Advertisement
Participants in our study positioned themselves not as enforcers, but as stewards of learning with integrity.
Their focus was on distinguishing between assistance that supports learning and assistance that substitutes for it. They identified three skill areas where assessment boundaries currently fall: prompting, critical thinking and writing.
Prompting: A legitimate and assessable skill
Participants widely viewed prompting – the ability to formulate clear and purposeful instructions for a chatbot – as a skill they could assess. Effective prompting requires students to break down tasks, understand concepts and communicate precisely.
Several noted that unclear prompts often produce poor outputs, forcing students to reflect on what they are really asking.
Advertisement
Prompting was considered ethical only when used transparently, drawing on one’s own foundational knowledge. Without these conditions, educators feared prompting may drift into over-reliance or uncritical use of AI.
Critical thinking
Educators saw strong potential for AI to support assessing critical thinking. Because chatbots can generate text that sounds plausible but may contain errors, omissions or fabrications, students must evaluate accuracy, coherence and credibility. Participants reported using AI-generated summaries or arguments as prompts for critique, asking students to identify weaknesses or misleading claims.
These activities align with a broader need to prepare students for work in a future where assessing algorithmic information will be a routine task. Several educators argued it would be unethical not to teach students how to interrogate AI-generated content.
Writing: Where boundaries tighten
Writing was the most contested domain. Educators distinguished sharply between brainstorming, editing and composition.
Advertisement
Brainstorming with AI was acceptable when used as a starting point, as long as students expressed their own ideas and did not substitute AI suggestions for their own thinking.
Editing with AI (for example, grammar correction) was considered acceptable only after students had produced original text and could evaluate whether AI-generated revisions were appropriate. Although some see AI as a legitimate support for linguistic diversity, as well as helping to level the field for those with disabilities or those who speak English as an additional language, others fear a future of language standardisation where the unique, authentic voice of the student is smoothed over by an algorithm.
Having chatbots draft arguments or prose was implicitly rejected. Participants treated the generative phase of writing as a uniquely human cognitive process that needs to be done by students, not machines.
Educators also cautioned that heavy reliance on AI could tempt students to bypass the “productive struggle” inherent in writing, a struggle that is central to developing original thought.
Advertisement
Our research participants recognised that in a hybrid cognitive future, skills related to AI, together with critical thinking are essential skills for students to be ready for the workforce after graduation.
Living in the post-plagiarism era
The idea of co-writing with GenAI brings us into a post-plagiarism era where AI is integrated into teaching, learning and communication in a way that challenges us to reconsider our assumptions about authorship and originality.
This does not mean that educators no longer care about plagiarism or academic integrity. Honesty will always be important. Rather, in a post-plagiarism context, we consider that humans and AI co-writing and co-creating does not automatically equate to plagiarism.
Today, AI is disrupting education and although we don’t yet have all the answers, it’s certain that AI is here to stay. Teaching students to co-create with AI is part of learning in a post-plagiarism world.
Advertisement
Design for a socially just future
Valid assessment in the age of AI requires clearly delineating which cognitive processes must remain human and which can be legitimately cognitively offloaded. To ensure higher education remains a space for ethical decision-making especially in terms of teaching, learning and assessment, we propose five design principles, based on our research:
Explicit expectations
The educator is responsible for making clear if and how GenAI can be used in a particular assignment. Students must know exactly when and how AI is a partner in their work. Ambiguity can lead to unintentional misconduct, as well as a breakdown in the student-educator relationship.
Process over product
By evaluating drafts, annotations and reflections, educators can assess the learning process, rather than just the output, or the product.
Design assessment tasks that require human judgement
Tasks requiring high-level evaluation, synthesis and critique of localised contexts are areas where human agency is still important.
Educators must teach students to be critical consumers of GenAI, capable of identifying its limitations and biases.
Preserving student voice
Assessments should foreground how students know what they know, rather than what they know.
Preparing students for a hybrid cognitive future
Educators in this study sought ethical, practical ways to integrate GenAI into assessment. They argued that students must understand both the capabilities and the limitations of GenAI, particularly its tendency to generate errors, oversimplifications or misleading summaries.
In this sense, post-plagiarism is not about crisis, but about rethinking what it means to learn and demonstrate knowledge in a world where human cognition routinely interacts with digital systems.
Advertisement
Universities and colleges now face a choice. They can treat AI as a threat to be managed, or they can treat it as a catalyst for strengthening assessment, integrity and learning. The educators in our study favour the latter.
Sarah Elaine Eaton is a professor and research chair for the Werklund School of Education, in the University of Calgary. Beatriz Antonieta Moya Figueroa is an assistant professor for the Werklund School of Education in the University of Calgary. Robert Brennan is a professor of mechanical and manufacturing engineering at the University of Calgary. Rahul Kumar is an assistant professor for the Faculty of Education at Brock University.
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.
European security officials believe two Russian space vehicles have intercepted the communications of at least a dozen key satellites over the continent. From a report: Officials believe that the likely interceptions, which have not previously been reported, risk not only compromising sensitive information transmitted by the satellites but could also allow Moscow to manipulate their trajectories or even crash them.
Russian space vehicles have shadowed European satellites more intensively over the past three years, at a time of high tension between the Kremlin and the West following Moscow’s full-scale invasion of Ukraine. For several years, military and civilian space authorities in the West have been tracking the activities of Luch-1 and Luch-2 — two Russian objects that have carried out repeated suspicious maneuvers in orbit.
Both vehicles have made risky close approaches to some of Europe’s most important geostationary satellites, which operate high above the Earth and service the continent, including the UK, as well as large parts of Africa and the Middle East. According to orbital data and ground-based telescopic observations, they have lingered nearby for weeks at a time, particularly over the past three years. Since its launch in 2023, Luch-2 has approached 17 European satellites.