Uladzislau Bayouski

A water meter buried in a concrete pit, three floors underground, sends one tiny reading a day. It has no power outlet, no Wi-Fi in range, and it needs to keep doing this for ten years on a battery you'll never replace. Wi-Fi can't reach it. Bluetooth and Thread can't travel that far. LoRaWAN could — if someone built and ran a gateway nearby. But there's already a network that blankets the entire city, penetrates deep underground, is professionally maintained, and charges pennies per device per month. It's the cellular network — running two special low-power modes built for exactly this. Those modes are NB-IoT and LTE-M . 📶 The Short Version NB-IoT and LTE-M sit in the gap no other protocol covers: the reach and reliability of the cellular network, stripped down to sip power and cost almost nothing. They're the licensed-spectrum answer to LoRaWAN — you don't run the infrastructure, a carrier does, and it already covers everywhere. NB-IoT is the...

Most people who own a Thread device have never heard of Thread. They bought an Eve sensor, a Nanoleaf bulb, or a Yale lock. It connected instantly. It responds in milliseconds. It still works when the internet goes down. They went back to living their life. That's Thread doing its job correctly — invisible, reliable, and fast enough to never be the reason something didn't work. 🏠 Thread vs OpenThread: Clear This Up First Two names, constant confusion. They are not the same thing. Thread is the standard — an open specification maintained by the Thread Group defining how low-power IPv6 mesh networks operate. It's the definition of what the network must do: self-healing mesh, IPv6-native, no single point of failure, AES-128 encryption end to end. OpenThread is the implementation — the open-source software stack, originally built by Google's Nest team, that runs the Thread specification on actual hardware. Thread is the recipe. OpenThread is the kitchen that m...

In the 1970s, a small group of engineers designing the internet made a decision that seemed absurdly generous at the time. They gave the network 4.3 billion addresses. Four billion. For a research project connecting a few dozen universities, that was infinity. Nobody in that room imagined a world where a single household would burn through forty of them — a phone, a laptop, a TV, a thermostat, a doorbell, a dozen bulbs, a watch, a speaker in every room. That world arrived. And when it did, the internet quietly ran out of room. 🌐 The Short Version IPv4 's 32-bit address space gives 4.3 billion unique identifiers. The central pool ran out in February 2011. Regional registries fell one after another: Asia-Pacific in April 2011, Europe in September 2012, Latin America in 2014, North America in September 2015. No fresh addresses left. The market responded the way markets do when something runs scarce: IPv4 addresses became an asset class. A single address that traded for ~$5 ...

You already own a Bluetooth mesh network. You just don't know it. If the lights in your office building dim automatically when the sun comes out, or the LED fixtures in a supermarket report their own energy use back to a dashboard, there's a decent chance a few thousand nodes are quietly relaying messages to each other over Bluetooth right above your head. Not the Bluetooth in your earbuds. Something fundamentally different. 📡 The Short Version Bluetooth mesh is a separate spec, built on BLE's radio but operating nothing like classic Bluetooth. BLE connects two devices — pair, hold a link, talk privately. Bluetooth mesh is a broadcast model: a device shouts, any node in earshot repeats it, that relay is heard by more nodes, who repeat it again. The message floods outward until it arrives. No pairing. No central coordinator. No center at all. The network scales as far as there are nodes to carry it — not as far as one radio can shout. Managed Flooding: The Be...

You buy a smart bulb. It needs its own app. You buy a smart lock from a different brand. That needs its own app too, and its own hub, and it doesn't show up in the first app at all. Your thermostat works with Google but not Apple. Your blinds work with Alexa only if you buy the bridge. Six apps. Three hubs. Nothing talks to anything. For fifteen years, that was the smart home. Not a platform — a pile. In 2026, that era is largely over. Two standards did it: Matter and Thread . 🏠 The Confusion Cleared Up First Matter and Thread are not competitors. They operate at different layers of the stack, and confusing them leads to bad buying decisions. Matter is the language — an application-layer standard defining how devices describe themselves, receive commands, report state, and authenticate. When a bulb says "I am a dimmable light, brightness 60%, range 0–100," it's speaking Matter. It runs on top of IP networks. Thread is the road — a low-power IPv6 mesh ...

Here's how it usually goes wrong. You build an IoT prototype. Fifty sensors, one reading a minute. You throw the data into PostgreSQL because that's what you know — and it works beautifully. Queries are instant. Everyone's happy. Then you go to production. Five thousand devices. One reading per second. Suddenly you're writing 5,000 rows every second — 432 million rows a day — and things start breaking in ways that feel personal. Dashboard queries that took 50ms now take 40 seconds. Your storage bill triples in a month. Adding an index makes writes worse. Removing it makes reads worse. Nothing is technically broken. You just picked the wrong tool, and IoT scale found the flaw. 📊 The Short Version IoT data has four properties that break traditional databases: Writes are relentless and append-only — high-frequency, continuous, never-ending; traditional B-tree indexes choke at sustained write rates Cardinality explodes at device scale — 10,000 devices × 5 se...

For decades, network security worked like a medieval castle. You built a big wall — the firewall. You put a gate in it — the VPN. Everything outside was dangerous. Everything inside was trusted. Once you were through the gate, you could roam freely, visit any room, open any door. That model is dead. And IoT is a big part of why it died. Fill your castle with thousands of cheap sensors, cameras, and controllers — many running outdated firmware, some with hardcoded passwords, most impossible to patch. Each one is a hole in your wall. When Mirai turned hundreds of thousands of them into an attack army, it proved the point brutally: the perimeter is meaningless when the threat is already inside, wearing a trusted uniform. 🔒 The Short Version Zero Trust rests on one foundational assumption: the network is already compromised. That sounds pessimistic. It's actually liberating. If you assume the attacker is already inside, you stop relying on the wall and start verifying ever...

Here's a threat that sounds like science fiction but is happening right now, while you read this. An attacker intercepts encrypted data from your IoT devices. They can't read it. The encryption is solid. So they don't try to break it. They just store it — copy the ciphertext to a drive and wait. They're betting that in five, eight, or ten years, a quantum computer will crack it retroactively. When that day comes, they'll read everything they harvested. The medical records. The industrial secrets. The authentication keys that might still be valid. This is called Harvest Now, Decrypt Later . It turns time itself into a weapon. A breach enabled in 2032 can originate from data intercepted in 2026. The clock has already started. 🔒 The Short Version The quantum threat is specific: Shor's algorithm can break RSA, ECC, and Diffie-Hellman — the public-key algorithms underpinning essentially all modern secure communication — on a sufficiently powerful quantum c...

For two decades, the deal with IoT was simple. Devices watched. They reported. A human read the alert and decided what to do. That model is being dismantled. The new generation doesn't wait for a human to read the alert. It detects the temperature spike, diagnoses the likely cause, adjusts the machine parameters to compensate, logs what it did, and only escalates if its own fix doesn't work. The human is no longer in the loop — they're watching over it. 🤖 The Short Version Autonomous IoT is the convergence of everything the IoT stack has been building toward: edge computing for real-time local processing, on-device AI for learned judgement, digital twins for outcome simulation, and lightweight protocols connecting it all. Put those pieces together and you get a device that doesn't just sense and report — it senses, reasons, decides, and acts. The distinction that matters most: automation vs autonomy. Automation follows fixed rules. "If temperature ...

You've shipped 50,000 smart locks across three continents. A security researcher emails on Tuesday afternoon. There's a buffer overflow in your firmware — exploitable, real, and present on every single lock. Without OTA, your options are recall, truck roll, or hope. Any of those costs millions and takes months. With OTA, you push a signed patch Tuesday evening. By Wednesday morning, 94% of your fleet is fixed. The remaining 6% update themselves when they next connect. That's why OTA isn't a feature. It's a survival mechanism. 🔒 The Short Version OTA (Over-the-Air) updates let you deliver new firmware to deployed IoT devices wirelessly — no physical access, no recall, no technician. The four components every OTA system needs: Update server — hosts firmware images, controls which devices get which version and when Device client — polls the server, downloads, verifies, and applies updates Transport layer — MQTT, HTTPS, or CoAP carrying the image secur...

A bearing on a production line starts to degrade. Nobody notices. The machine keeps running. Three days later it fails catastrophically — the line stops, repair parts are on a six-week lead time, a full shift of output is gone. Now replay that scenario with predictive maintenance running. The vibration sensor notices the shift on day one. The AI flags early-stage bearing degradation. A work order is automatically generated. The bearing is replaced during a scheduled lunch break on day two. The production line never stops. That's not a hypothetical. That's happening right now in factories, power plants, aircraft fleets, and wind farms worldwide. 🔧 The Short Version Most maintenance is either reactive (fix it when it breaks — expensive chaos) or preventive (service on a schedule — up to 30% of work is unnecessary). Predictive maintenance replaces both with a third approach: service it when the data says it needs it. The economics are hard to argue with: 30–50% reductio...