The hunt is on for anything that can surmount AI’s perennial memory wall–even quick models are bogged down by the time and energy needed to carry data between processor and memory. Resistive RAM (RRAM)could circumvent the wall by allowing computation to happen in the memory itself. Unfortunately, most types of this nonvolatile memory are too unstable and unwieldy for that purpose.

Fortunately, a potential solution may be at hand. At December’s IEEE International Electron Device Meeting (IEDM), researchers from the University of California, San Diego showed they could run a learning algorithm on an entirely new type of RRAM.

“We actually redesigned RRAM, completely rethinking the way it switches,” says Duygu Kuzum, an electrical engineer at the University of California, San Diego, who led the work.

RRAM stores data as a level of resistance to the flow of current. The key digital operation in a neural network—multiplying arrays of numbers and then summing the results—can be done in analog simply by running current through an array of RRAM cells, connecting their outputs, and measuring the resulting current.

Traditionally, RRAM stores data by creating low-resistance filaments in the higher-resistance surrounds of a dielectric material. Forming these filaments often needs voltages too high for standard CMOS, hindering its integration inside processors. Worse, forming the filaments is a noisy and random process, not ideal for storing data. (Imagine a neural network’s weights randomly drifting. Answers to the same question would change from one day to the next.)

Moreover, most filament-based RRAM cells’ noisy nature means they must be isolated from their surrounding circuits, usually with a selector transistor, which makes 3D stacking difficult.

Limitations like these mean that traditional RRAM isn’t great for computing. In particular, Kuzum says, it’s difficult to use filamentary RRAM for the sort of parallel matrix operations that are crucial for today’s neural networks.

So, the San Diego researchers decided to dispense with the filaments entirely. Instead they developed devices that switch an entire layer from high to low resistance and back again. This format, called “bulk RRAM”, can do away with both the annoying high-voltage filament-forming step and the geometry-limiting selector transistor.

The San Diego group wasn’t the first to build bulk RRAM devices, but it made breakthroughs both in shrinking them and forming 3D circuits with them. Kuzum and her colleagues shrank RRAM into the nanoscale; their device was just 40 nm across. They also managed to stack bulk RRAM into as many as eight layers.

With a single pulse of identical voltage, an eight-layer stack of cells each of which can take any of 64 resistance values, a number that’s very difficult to achieve with traditional filamentous RRAM. And whereas the resistance of most filament-based cells are limited to kiloohms, the San Diego stack is in the megaohm range, which Kuzum says is better for parallel operations. e

“We can actually tune it to anywhere we want, but we think that from an integration and system-level simulations perspective, megaohm is the desirable range,” Kuzum says.

These two benefits–a greater number of resistance levels and a higher resistance–could allow this bulk RRAM stack to perform more complex operations than traditional RRAM’s can manage.

Kuzum and colleagues assembled multiple eight-layer stacks into a 1-kilobyte array that required no selectors. Then, they tested the array with a continual learning algorithm: making the chip classify data from wearable sensors—for example, reading data from a waist-mounted smartphone to determine if its wearer was sitting, walking, climbing stairs, or taking another action—while constantly adding new data. Tests showed an accuracy of 90 percent, which the researchers say is comparable to the performance of a digitally-implemented neural network.

This test exemplifies what Kuzum thinks can especially benefit from bulk RRAM: neural network models on edge devices, which may need to learn from their environment without accessing the cloud.

“We are doing a lot of characterization and material optimization to design a device specifically engineered for AI applications,” Kuzum says.

The ability to integrate RRAM into an array like this is a significant advance, says Albert Talin, materials scientist at Sandia National Laboratories in Livermore, California, and a bulk RRAM researcher who wasn’t involved in the San Diego group’s work. “I think that any step in terms of integration is very useful,” he says.

But Talin highlights a potential obstacle: the ability to retain data for an extended period of time. While the San Diego group showed their RRAM could retain data at room temperature for several years (on par with flash memory), Talin says that its retention at the higher temperatures where computers actually operate is less certain. “That’s one of the major challenges of this technology,” he says, especially when it comes to edge applications.

If engineers can prove the technology, then all types of models may benefit. This memory wall has only grown higher this decade, as traditional memory hasn’t been able to keep up with the ballooning demands of large models. Anything that allows models to operate on the memory itself could be a welcome shortcut.

After launching a “Home Safe” feature that lets users notify friends and family when they’ve arrived home safely, Snapchat is now introducing additional alerts to inform others when users have arrived at other destinations.

The social media giant announced on Monday that with its new “Arrival Notifications,” users can now set one-time or recurring alerts for locations beyond their home, providing an automatic way to share when they’ve arrived at specific places.

“Arrival Notifications now work for everyday moments — like letting someone know you’re back for the night while traveling, or automatically sharing when you arrive at a weekly class, practice, or meeting — without needing to remember to send a message,” the company wrote in a blog post.

As with the platform’s Home Safe alerts, Arrival Notifications can only be sent to friends you choose to share your location with. It’s worth noting that location sharing on Snap Map is off by default. No one can see your location or receive an alert unless you choose to share it, Snapchat explained. One-time alerts expire after they’re sent or after 24 hours.

To use Arrival Notifications, you need to share your location with a trusted friend that you want to keep in the loop. Then, you need to tap on your friendship profile and scroll down to “Arrival Notifications.” You can pick a location on the map and give it a personal name. For example, you could set the location for your “run club” or the location for “piano lessons.” You can then choose a one-time or recurring alert, after which Snapchat will notify your friend when you arrive.

The new feature comes as Snapchat announced last summer that Snap Map now has more than 400 million monthly active users.

Snap Map, which launched in 2017, was originally a way for users to see their friends’ locations and browse public snaps from around the world. The feature now also offers ways for users to discover local hotspots and find things to do.

With its Home Safe and Arrival Notifications features, Snapchat is looking to further compete with services like the family location-sharing app Life360 and Apple’s “Find My.”

In the early days of AI, a common example program was the hexapawn game. This extremely simplified version of a chess program learned to play with your help. When the computer made a bad move, you’d punish it. However, people quickly realized they could punish good moves to ensure they always won against the computer. Large language models (LLMs) seem to know “everything,” but everything is whatever happens to be on the Internet, seahorse emojis and all. That got [Hayk Grigorian] thinking, so he built TimeCapsule LLM to have AI with only historical data.

Sure, you could tell a modern chatbot to pretend it was in, say, 1875 London and answer accordingly. However, you have to remember that chatbots are statistical in nature, so they could easily slip in modern knowledge. Since TimeCapsule only knows data from 1875 and earlier, it will be happy to tell you that travel to the moon is impossible, for example. If you ask a traditional LLM to roleplay, it will often hint at things you know to be true, but would not have been known by anyone of that particular time period.

Chatting with ChatGPT and telling it that it was a person living in Glasgow in 1200 limited its knowledge somewhat. Yet it was also able to hint about North America and the existence of the atom. Granted, the Norse apparently found North America around the year 1000, and Democritus wrote about indivisible matter in the fifth century. But that knowledge would not have been widespread among common people in the year 1200. Training on period texts would surely give a better representation of a historical person.

The model uses texts from 1800 to 1875 published in London. In total, there is about 90 GB of text files in the training corpus. Is this practical? There is academic interest in recreating period-accurate models to study history. Some also see it as a way to track both biases of the period and contrast them with biases found in data today. Of course, unlike the Internet, surviving documents from the 1800s are less likely to have trivialities in them, so it isn’t clear just how accurate a model like this would be for that sort of purpose.

Instead of reading the news, LLMs can write it. Just remember that the statistical nature of LLMs makes them easy to manipulate during training, too.

Featured Art: Royal Courts of Justice in London about 1870, Public Domain

Lukasz Swiatek of the University of New South Wales Sydney discusses what advancement in technologies might mean for future graduates.

The head of the International Monetary Fund, Kristalina Georgieva, has warned young people will suffer the most as an AI “tsunami” wipes out many entry-level roles in coming years.

Tasks that are eliminated are usually what entry-level jobs do at present, so young people searching for jobs find it harder to get to a good placement.

Georgieva is not alone. Other economic and business experts have warned about AI taking entry-level jobs.

As young people prepare to start or continue their university studies, they may be feeling anxious about what AI means for their job prospects. What does the current research say? And how can you prepare for a post-AI workforce while studying?

The situation around the world

At the moment, the impact of AI is uneven and depends on the industry.

A 2025 report from US think tank the Brookings Institution suggests, in general, AI adoption has led to employment and firm growth. Most importantly, AI has not led to widespread job loss.

At the same time, consulting firm McKinsey notes many businesses are experimenting with AI and redesigning how they work. So, some organisations are seeking more technically skilled employees.

Crucially, AI is affecting each industry differently. So, we might see fewer entry-level jobs in some industries, but more in others, or growth in specialist roles.

For example, international researchers have noted agriculture has been a slow adopter of AI. By contrast, colleagues and I have found AI is being rapidly implemented in media and communications, already affecting jobs from advertising to the entertainment industries. Here we are seeing storyboard illustrators, copywriters and virtual effects artists (among others) increasingly being replaced by AI.

So, students need to look carefully at the specific data about their chosen industry (or industries) to understand the current situation and predicted trends.

To do this, you can look at academic research about AI’s impacts on industries around the world, as well as industry news portals and free industry newsletters.

Get ready while studying

Students can also obviously build their knowledge and skills about AI while they are studying.

Specifically, students should look to move from “AI literacy” to “AI fluency”. This means understanding not just how AI works in an industry, but also how it can be used innovatively in different contexts.

If these elements are not already offered by your course, you can look at online guides and specific courses offered by universities, TAFE or other providers.

Students who are already familiar with AI can keep expanding their knowledge and skills. These students can discover the latest research from the world’s key publishers and keep up to date with other AI research news.

For students who aren’t really interested in AI, it’s still important to start getting to grips with the technology. In my research, I’ve suggested getting curious initially about three key things: opportunities, concerns and questions. These three elements can be especially helpful for getting across industry developments: how AI is being used, what issues it’s raising, and which impacts still need to be explored.

Free (online) courses, such as AI For Everyone and the Elements of AI, can help familiarise virtually anyone with the technology.

Strengthening other skills

All students, no matter how familiar they are with AI, can also concentrate on developing general competencies that can apply across any industry. US researchers have pinpointed six key “durable skills” for the AI age:

• effective communication, to engage with others successfully
• good adaptability, to respond to workplace, industry and broader social changes
• strong emotional intelligence, to help everyone thrive in a workplace
• high-quality creativity, to work with AI in innovative ways
• sound leadership, to help navigate the challenges that AI creates
• robust critical thinking, to deal with AI-related problems.

So, look for opportunities to foster these skills in and out of class. This could include engaging in teamwork, joining a club or society, doing voluntary work, or getting paid work experience.

Don’t forget ethics

Finally, students need to consider the ethical issues this new technology creates. Research suggests AI is bringing about changes in ethics across industries and students need to know how to approach AI dilemmas.

For example, they need to feel confident tackling questions about when to use and not use AI, and whether the technology’s environmental impacts outweigh its benefits in different situations.

Students can do this through focused discussions with classmates, facilitated by teachers to tease out the issues. They can also do dedicated courses on AI ethics.

By Lukasz Swiatek

Lukasz Swiatek lectures in the School of the Arts and Media at UNSW Sydney. His main research areas are media and communication, higher education, and cultural studies. Over the years, he has taught a range of postgraduate and undergraduate courses, in media studies, communication, international and global studies.

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.

The BBC recently published an exposé revealing that some Chinese subscription sites charge for access to their network of hundreds of hidden cameras in hotel rooms. Of course, this is presumably without the consent of the hotel management and probably isn’t specifically a problem in China. After all, cameras can now be very tiny, so it is extremely easy to rent a hotel room or a vacation rental and bug it. This is illegal, China has laws against spy cameras, and hotels are required to check for them, the BBC notes. However, there is a problem: At least one camera found didn’t show up on conventional camera detectors. So we wanted to ask you, Hackaday: How do you detect hidden cameras?

How it Works

Commercial detectors typically use one of two techniques. It is easy to scan for RF signals, and if the camera is emitting WiFi or another frequency you expect cameras to use, that works. But it also misses plenty. A camera might be hardwired, for example. Or store data on an SD card for later. If you have a camera that transmits on a strange frequency, you won’t find it. Or you could hide the camera near something else that transmits. So if your scanner shows a lot of RF around a WiFi router, you won’t be able to figure out that it is actually the router and a small camera.

The other common method uses a beam of light or a laser to try to see reflections of lenses, which will be retroreflective. The user views the room through a viewfinder, and any light that comes directly back will show up in the view. Despite some false positives, this method will find cameras even if they are not powered or transmitting. Even shining a flashlight, maybe from the same cell phone, around a dark room might uncover some camera devices.

There are a few other techniques. If you assume a spy camera probably uses IR lighting to see you at night, you can scan for that. A good tip is that your cell phone camera can probably see IR. (Test it on an IR remote control.) So looking around with your phone camera is a good, free way to find some cameras. A thermal imager might show hidden equipment, too, although it might be hard to determine if it is actually a camera or not.

You might be thinking: just look for the camera. But that’s not always simple. In the BBC article, the camera was the size of a pencil eraser. Not to mention, a quick search of your favorite retailer will reveal cameras made to look like smoke detectors, stuffed toys, USB chargers, and more. You can even get small cameras that can mount a fake button or screw head on the lens.

Testing

[Project Farm] has a video that tests a few detectors. The problem, of course, is that there are different kinds of cameras. Detecting the test camera doesn’t mean it will detect all cameras. Still, you can get some idea of how effective some detectors are compared to others.

Your Turn?

Given that none of the current ways to detect cameras work perfectly, what would you build to find them? Maybe an NLJD? Or maybe some tech to blind them? Tell us what you think in the comments.

In Mr. Seevers’ English class, the air feels different today. A quiet student draws an unexpected connection between “The Odyssey” and modern migrant stories. The room wakes up. One idea sparks another, and conversation bounces around the room. Mr. Seevers grins, scribbling connections on the whiteboard, forgetting the clock. Students lean in. Their voices matter. Teacher and students move together — focused, curious, absorbed. When the bell finally rings, Mr. Seevers realizes … this is why he never quit.

Moments like this still happen in classrooms, but maybe not often enough to sustain the enthusiasm most educators had when they started out. As a result, leaders are grappling with two familiar challenges: finding and keeping great teachers. Seats go unfilled, turnover disrupts continuity and costs balloon.

But more troubling is the quiet toll that teacher burnout takes on students. In an era of artificial intelligence, shifting expectations and high-stakes accountability, students need teachers who are present, enthusiastic, resilient and growth-minded.

The key to recruiting and retaining great teachers is helping them find and sustain a sense of “flow” — a state where their energy, purpose and performance align — and bring that vitality into classrooms. While burnout depletes a teacher’s psychic energy, flow replenishes it. The payoff for districts is profound: stronger retention, smoother recruitment and better student outcomes.

What Is Flow, and Why It Matters in the Classroom

Flow is the state of full absorption and intrinsic motivation. It’s the sweet spot between boredom (too easy) and anxiety (too hard). It’s that “in the zone” moment where “everything just clicks.”

In K-12 education, flow tends to center around student learning: creating lessons that challenge just enough, focusing attention without overwhelming it. But teachers benefit from that psychological sweet spot, too: planning in flow, teaching in flow and iterating in flow.

When teachers find flow, something subtle but powerful happens. Their focus and curiosity become contagious. Research on emotional contagion shows that a teacher’s mood shapes the climate of the classroom. Stress and frustrations spread quickly, but so do calm, curiosity and joy. Most teachers don’t realize how strongly their inner state influences student engagement, but it does.

Flow feeds on itself. The more teachers experience it, the more students do, creating cycles of focus, persistence and connection that drive better outcomes.

How the Pygmalion Effect Fuels Flow

The Pygmalion effect, also known as the Rosenthal effect, describes how higher expectations from teachers can lead to improved student performance. Educators communicate those expectations through tone, feedback, time and warmth, cues that shape how students view themselves as learners.

Those same beliefs drive teacher flow. While low expectations lead to overly easy lessons and unrealistically high expectations produce anxiety, teachers who believe their students can grow naturally design learning that stretches skills without overwhelming them, the ideal balance for flow.

That energy is contagious, too. Studies indicate that teacher flow can cross over to students, creating a “flow contagion,” an upward spiral of shared engagement and persistence. The Pygmalion effect sets the stage; flow helps bring it to life.

The Recruitment and Retention Problem, and Why Mindset Matters

Districts today are busy chasing recruitment metrics (number of applicants, credentialing pipelines), but retention is where the real crisis lies. K-12 teachers now report the highest burnout rates in the country, across all jobs and industries. Systemic pressures like inadequate funding, excessive workloads, challenging student behaviors, parent scrutiny and lack of administrative support have them constantly on their back foot instead of in their zone of flow.

Consider a school district rolling out new AI tools to streamline lesson planning, grading or student feedback. On paper, it’s a practical move to save time and modernize instruction. But without coaching or dialogue, many teachers feel blindsided. They’re being asked to integrate complex tools while juggling a full teaching load, testing demands and students’ social-emotional needs.

The result? Instead of embracing the technology and feeling empowered, teachers feel alienated. Some worry that creativity, intuition and human connection matter less than AI adoption. Those already close to burnout may see AI as one more way their professional judgment is being replaced or devalued.

That stress doesn’t stay contained. It seeps into the classroom. Students pick up on frustration and unease just as easily as they absorb enthusiasm and curiosity.

When AI integration is paired instead with coaching and the spirit of exploration, teachers have space to process fears, experiment with tools and reflect on what works. They move from compliance to curiosity. This relational support can transform AI from a threat into a trusted collaborator, helping educators reclaim time, creativity and joy in their work.

The Benefits of Coaching and Teacher Flow

Districts that invest in coaching can strengthen multiple points in their teacher lifecycle:

• Onboarding: New teachers get guidance on finding flow in their planning and instruction.
• Burnout prevention: Coaches help identify stressors early, redesign workflows and create guardrails for energy.
• Sustained engagement: Teachers experience coaching as developmental support from their district.
• Improve recruiting: Prospective hires see a workplace that values professional well-being.

When retention improves, districts recover not only in lowered costs but in preserved institutional memory, relationships, curricular coherence and, most critically, consistently high instructional quality.

How Teacher Flow Translates to Student Outcomes

• Sustained energy → better pacing: Teachers who maintain flow teach more responsively, observe more closely and adjust in the moment.
• Mindset-aligned expectations → higher student growth: Teachers who believe in student potential push appropriately, scaffold growth and persist.
• Emotional contagion → classrooms that hum: A teacher in flow models calm, curiosity and agency, and students respond in kind.
• Upward spirals of engagement: As students engage, teachers get feedback, adapt and reenter flow.
• Reduced classroom disruption: Lower turnover means fewer substitutes, fewer gaps and more continuity.

Prioritize Human-Centered Support Systems

Administrators don’t control every budget or class-size metric, but they can decide how people are supported, how leaders lead and how change takes shape. The difference between a district that churns teachers and one that nurtures them often comes down to access to coaching, a growth mindset, relational support and an environment that values energy, flow and reflective practice.

Districts that cultivate a coaching culture can move from perpetual retention crisis to a thriving educational community where teachers and students grow together. Learn more about how BetterUp supports educator well‑being.