Montreal, August 10, 2026 — Welcome back to Tech Science Daily, PcHybrid's plain-language tour of the science behind this week's technology headlines. The past seven days delivered an unusually rich crop: Samsung put a microscopic red-green-blue LED array behind a television panel and expanded it into a full lineup, memory engineers sketched a future where DRAM sits physically on top of the processor, and a self-replicating worm chewed through some of the most-downloaded packages in the JavaScript ecosystem. Below you'll find our ten-story radar, then three deep dives that explain the physics, the engineering trade-offs, and — because we are a store as well as a science desk — what it all means for the hardware you might actually buy. As always, every product we recommend was checked against our live inventory this morning.

Today's Tech Radar

Deep Dive 1 — Micro RGB: What Happens When the Backlight Learns to Paint

Every LCD television you have ever owned paints its picture with borrowed light. Micro RGB changes who holds the brush. Photo: Jonas Leupe / Unsplash.

The science: from white floodlight to a million tiny paintbrushes

To understand why Samsung's Micro RGB announcement matters, start with how a conventional LCD TV works. The liquid-crystal panel you look at does not emit light at all — it is a shutter. Behind it sits a backlight, historically a sheet of white or blue LEDs whose light passes through colour filters to become the reds, greens and blues of the picture. This design has two built-in compromises. First, the colour filters are subtractive: they throw away most of the light to isolate each primary colour, wasting energy and limiting how saturated colours can be. Second, the backlight is coarse. Even a good Mini LED TV with a few thousand dimming zones is trying to describe an eight-million-pixel image with a few thousand patches of brightness, which is why bright objects on dark backgrounds can "bloom" with a halo of stray light.

Micro RGB attacks both compromises at once. According to Samsung's announcement, the backlight is built from individually controlled red, green and blue LEDs, each less than 100 micrometres across — roughly the width of a human hair — arranged in an ultra-fine grid behind the panel. Because the light source itself is now coloured, the display no longer relies on filters to create its primaries; it mixes red, green and blue light directly, the same additive principle your eye's cone cells use. And because each micro-LED is individually addressable, the dimming zones shrink from patches to points. The practical results Samsung claims are striking: coverage of 100 percent of the BT.2020 colour gamut — the widest standard used in HDR mastering, which no consumer LCD had fully covered — and near broadcast-monitor colour accuracy, managed frame-by-frame by a dedicated "Micro RGB AI Engine Pro" processor that analyses each scene and adjusts the array in real time.

It helps to place Micro RGB on the family tree of display technology. OLED puts the light-emitter in every pixel (perfect black, but limited peak brightness and organic materials that age). MicroLED — the technology in Samsung's wall-sized luxury displays — puts inorganic emitters in every pixel, but manufacturing millions of transferable microscopic LEDs per screen remains ferociously expensive. Micro RGB is a pragmatic middle path: it keeps the mature, high-yield LCD panel for resolution, and moves the innovation into the backlight, where tens of thousands of micro-LEDs are enough. You get most of the colour-purity and local-contrast benefit at a fraction of the manufacturing risk. That is why Samsung could launch the world-first 115-inch model at $29,999.99 US and then, within months, announce a 2026 lineup spanning 55, 65, 75, 85, 100 and 115 inches — a classic case of a halo technology beginning its march down the price ladder.

Industry context and what to buy today

For buyers, the lesson from every previous display transition (LED backlights, quantum dots, Mini LED) is the same: year-one flagship pricing falls fast, and the mainstream models you can buy today remain excellent value precisely because the spotlight has moved. If you need serious screen real estate now — for a boardroom, a storefront window, a lobby or a very ambitious home theatre — commercial-grade 4K panels deliver brightness, anti-glare coatings and 24/7 duty ratings that consumer TVs don't match. In our own warehouse, the Samsung QM85C 85-inch UHD display (in stock) offers 500-nit output with a non-glare finish and an IP5X dust rating, while the more compact Samsung 55" Crystal UHD Signage QBC (in stock) brings the same slim commercial design to smaller spaces at a far friendlier price. For desk-scale colour work — photo editing, video grading, or simply enjoying 4K content up close — the Samsung ViewFinity S8 32" 4K monitor (in stock) covers the essentials of accurate colour without flagship-TV money. Not sure which panel size or brightness class fits your room and lighting? You can always request a free quote from our team and we'll spec it with you.

Deep Dive 2 — zHBM and the Memory Wall: Why the Next Leap in Computing Is Vertical

Processors have grown faster far quicker than the wires feeding them data. Stacking memory on top of the chip shortens the commute to micrometres. Photo: Alexandre Debiève / Unsplash.

The science: when the bottleneck is the wire, not the transistor

At the Future of Memory and Storage summit (FMS 2026) this week, Samsung showed a concept it calls zHBM: high-bandwidth memory stacked directly on top of an AI accelerator, rather than sitting beside it. To see why engineers are excited, you need one idea: the memory wall. For decades, processor performance has grown faster than the speed at which data can be moved into the processor. A modern AI chip can perform trillions of operations per second, but each operation needs operands fetched from memory — and moving a bit of data across a few centimetres of circuit board costs more energy, and often more time, than the arithmetic performed on it. Large language models make this brutal: their weights are hundreds of gigabytes, and generating each token means streaming enormous tensors through the chip. In many real workloads, the expensive compute units simply sit idle, starved by the interconnect.

High Bandwidth Memory (HBM) was the first fix: stack DRAM dies vertically, connect them with through-silicon vias (TSVs) — microscopic copper elevators drilled through the silicon — and place the stack next to the processor on a shared silicon interposer. That shortened the data commute from centimetres to millimetres. zHBM proposes the logical endpoint: eliminate the sideways trip entirely by bonding the memory stack on top of the compute die, so signals travel micrometres, straight down. Shorter wires mean lower capacitance, which means less energy per bit and higher achievable frequencies; wafer-to-wafer hybrid bonding also allows far more connections per square millimetre than solder bumps. The claimed numbers from TrendForce's and Samsung's presentations are dramatic: roughly eight times the performance of HBM5 (itself still a future standard), more than ten times the memory density, about three times the energy efficiency, and thermal resistance cut by more than half — the last point being crucial, because the historical objection to piling memory on a hot processor was always heat. Alongside zHBM, Samsung previewed zNAND-O, its implementation of the emerging High Bandwidth Flash (HBF) idea, which stacks NAND flash near the processor to provide a huge, slower-but-cheaper tier of nearby capacity, plus 400-plus-layer bonded NAND and HBM4E samples. The target window for zHBM is around 2029, and it remains a concept — but SK Hynix presented its own vertical-stacking ideas the same week, and when the two companies that dominate the DRAM market agree on a direction, the industry tends to follow.

Why a data-centre technology matters to your next laptop

The memory hierarchy in your own devices already follows the same logic in miniature: the closer the storage sits to the processor, the faster everything feels. That's why today's best laptops pair on-package memory with NPUs (neural processing units) for local AI, and why a fast NVMe SSD transforms an ageing desktop. If you're buying now, an AI-ready ultraportable such as the Asus ExpertBook P5 with Intel Core Ultra 5 (in stock) uses Intel's Lunar Lake-class design, where memory sits on the processor package for exactly the latency and power reasons described above. Prefer ARM efficiency? The Lenovo ThinkPad T14s Gen 6 with Snapdragon X Plus (in stock) delivers a 45-TOPS NPU and all-day battery from the same system-on-chip philosophy. For desktops, the cheapest taste of the "shorter wires" principle is storage: the Samsung 990 PRO 1TB NVMe SSD (in stock) rides PCIe Gen4 directly to the CPU rather than detouring through a SATA controller. And on-device AI is not just a PC story — the Samsung Galaxy S25 Ultra (in stock) runs Galaxy AI features locally on its Snapdragon 8 Elite silicon precisely because its memory subsystem is fast enough to feed the NPU, and the Galaxy Tab S10 FE (in stock) brings the same on-device intelligence to a 10.9-inch canvas.

Deep Dive 3 — Shai-Hulud Returns: Anatomy of a Self-Replicating Supply-Chain Worm

Supply-chain attacks don't break your lock — they arrive already inside the package you invited in. Photo: FlyD / Unsplash.

The science: how a worm spreads through a package registry

The week's biggest security story is the return of Shai-Hulud, a self-propagating worm targeting npm, the package registry that underpins most of the JavaScript world. On August 4, researchers at Elastic Security Labs identified a new campaign — dubbed CHAINDROP — that compromised the maintainer of keyv, a widely used key-value storage library, and trojanized more than 400 packages. Advisories from national cyber agencies and multiple security firms put the cumulative campaign at over 1,300 compromised package versions whose legitimate counterparts represent roughly two billion downloads per month, including staples like keyv, cacheable, flat-cache and file-entry-cache.

What makes a worm different from ordinary malware is autonomy: it spreads without further attacker involvement, and understanding the loop explains why registries are such fertile ground. Step one: the malicious code hides in a package's preinstall script, which npm executes automatically the moment anyone — a developer, a build server, an automated dependency updater — installs the package, before any tests or security checks run. Step two: on execution it harvests every credential it can find on the machine: npm publishing tokens, GitHub tokens, cloud keys. Step three is the reproductive trick: using the stolen npm token, it automatically publishes backdoored versions of every other package the victim has publish rights to. Each newly infected package recruits its own downloaders as the next generation of hosts — exponential spread, in the same mathematical family as biological epidemics. In a final touch of cruelty, the malware creates a public GitHub repository named "Shai-Hulud" under the victim's account and commits their stolen secrets to it, exposing them to the entire internet. This wave arrived alongside a broader trend the weekly threat roundups keep flagging: new passkey-abuse techniques hijacking accounts, and actively exploited flaws in Chrome, Cisco IOS XE, Jenkins, TeamCity and Veeam. Attackers concentrate on identity, automation and human error because those routes are cheaper than breaking hardened infrastructure.

What a small business (or a careful individual) should actually do

You don't need to be a software company to be exposed — if your website, point-of-sale, or accounting stack was built with modern web tooling, npm is somewhere in your supply chain. The practical defences are refreshingly concrete. Keep operating systems and browsers current, since several of this week's exploited bugs were patched weeks ago. Treat any developer machine or build server as compromised if it installed affected packages, and rotate its credentials. Most importantly, move your critical logins to phishing-resistant authentication: hardware-backed FIDO2 keys mean a stolen password (or a replayed session) is no longer enough. We stock the Kensington VeriMark Guard USB-C fingerprint security key (in stock), which implements FIDO2/WebAuthn with a biometric touch — the same class of protection that would have blunted many of this year's identity attacks. Segmenting risky work also helps: doing your banking and admin on a clean, dedicated device such as the affordable Samsung Galaxy Tab A9+ (in stock) keeps your most sensitive sessions away from machines that install software all day. If you'd like an expert eye on your setup — from password-manager rollout to hardware-key deployment for a whole team — request a free quote from our team and we'll map out a security refresh that fits your budget.

Glossary of the Week

Setup at a Glance

That's the science for today. If any of this sparked a project — a display wall, an AI-ready laptop fleet, or a long-overdue security refresh — request a free quote from our team and we'll turn the physics into a shopping list. See you tomorrow.

Sources & Further Reading

Samsung Newsroom — World-first Micro RGB TV · Samsung Newsroom — 2026 Micro RGB lineup expansion · ecoustics — 2026 Micro RGB pricing and availability · Samsung Newsroom — Next-gen 3D memory vision at FMS 2026 · TrendForce — Samsung showcases zHBM at FMS 2026 · SamMobile — zHBM concept, 8x HBM5 · Tom's Hardware — Co-packaged optics and next-gen memory week in review · Elastic Security Labs — CHAINDROP hits 400+ npm packages · Cyber Security Agency of Singapore — Shai-Hulud advisory (keyv and related packages) · Wiz — keyv and cacheable hijacked · Akamai — the worm returns and goes public · CyberPress — Weekly cybersecurity roundup, Aug 3–7, 2026 · TechStartups — AMD record Q2 and the week's AI hardware news · TechRadar — August 2026 hardware roundup (LG panel, Acer Swift Go). Photos: Unsplash (free commercial license).