Substack started life as a way for writers to send paid newsletters to people who actually wanted them. The app that carries the name now does considerably more than deliver email.

Listed under News & Magazines and published under the name Substack Inc., it describes itself as a media app built around creators, ideas, and communities, and the feed reflects that: essays sit next to video episodes, podcasts, short clips, and live group chats.

For anyone who has spent years bouncing between a podcast app, a video platform, and an inbox full of half-read newsletters, the appeal is the consolidation. One place, one list of subscriptions, one set of notifications.

Beyond the newsletter

The core unit on Substack is still a publication, and behind each one is an independent writer, artist, or podcaster rather than an editorial board. That structure shapes what you find. The catalogue spans food, sports, politics, fashion, comedy, and finance, and the tone tends to be more opinionated than what a general news app serves up, because there is no house style to conform to.

Some publications are text-first with an occasional audio version. Others are video shows first, with written notes attached almost as an afterthought. Serialized writing does well here too, with fiction and long-running project diaries arriving in installments the way a column once did.

Free and paid tiers

Subscriptions come in two flavours. You can follow a publication for free and read whatever the creator makes public, or pay to read the subscriber-only work and, in most cases, get access to comment threads and direct contact with the writer.

That paid layer is the whole economic engine of the platform, which is why creators talk about it so much. From a reader’s side it means the decision is per publication, not per platform. You can carry a dozen free follows and one paid subscription without any of the bundling that streaming services push.

The app also keeps videos and podcasts ad-free, including read-aloud versions of articles, which is a meaningful difference if your podcast queue currently opens with ninety seconds of mattress commercials.

When writers go live

The live features are the part people tend to underestimate. Creators host livestreams and live group chats, usually for subscribers, and these are closer to a call-in radio show than a comment section. A finance writer walking through a market selloff in real time, or a comedy podcaster taking questions after an episode, works better as an event than as a post.

Chat threads sit somewhere between a forum and a group message. They stay attached to a publication, so the conversation tends to keep its shape instead of drifting the way open social feeds do.

Substack on Windows

Because so much of the reading here is long, a desktop screen suits the material. Substack is an Android app rather than a Windows program, so running it on a PC means using an Android emulator such as BlueStacks. Some people prefer LDPlayer for its lighter footprint, though either approach puts the same mobile app in a resizable window next to your browser.

Two things tend to trip people up. Audio playback in an emulator can behave oddly if the virtual machine grabs the wrong output device, which matters when a large share of what you consume is podcasts and read-aloud posts. Livestreams also want a stable connection and hardware acceleration turned on, or video will stutter while the chat keeps scrolling.

If either of those snags hits, or you just want to confirm things run cleanly before committing to a paid subscription you plan to read at your desk, Substack for Windows walks through fixing the audio routing problem and getting the Android build running in a desktop window. Signing in with the same account you use on your phone keeps your subscription list and reading position in sync.

Settling in as a subscriber

Notifications are the setting worth visiting first. A dozen active follows across text, audio, and live events will generate a lot of pings, and the per-publication controls are more useful than the global toggle.

Downloading audio for offline listening is handy on a phone and less so on a desktop, though it does mean an episode keeps playing if your connection drops mid-stream. Search rewards specificity: browsing by category surfaces the big names, while searching a niche subject tends to turn up the smaller publications that made the platform interesting in the first place.

If you do subscribe to a finance writer, keep the livestream schedule handy. Watching someone talk through a selloff as it happens is a different experience from reading the recap the next morning.

Local-language input is one of those problems that looks solved until you need it. Latin-script languages get keyboard support by default on every operating system, but scripts like Sinhala sit in an awkward middle ground: the fonts render fine, Unicode support exists, yet the act of typing remains clumsy without software built for the language rather than bolted onto it.

People in this space usually reach for one of three things. A transliteration tool that turns romanized text into the target script, a system-level keyboard layout that assumes you already know where every character lives, or a browser-based converter you paste into and copy out of. Each works. Each also breaks the moment you want to type a quick reply without switching contexts.

That gap is where dedicated local-language keyboard apps have carved out real user bases, and in Sri Lanka the clearest example is Helakuru.

Four million daily typists and a trophy shelf

Helakuru (හෙළකුරු) is listed under Lifestyle rather than Tools, which sounds odd for a keyboard until you look at the scale involved. The app describes itself as Sri Lanka’s number one app with over four million users, and its award list backs up the reach: Most Popular App of Sri Lanka at the NYDB Awards 2016, the mBillionth South Asian Award in 2014, an e-Swabhimani National Award in 2013, an NBQSA National Award in 2014, and Sri Lanka’s Best Local Language Product Award the same year.

Awards age quickly. The user number is the more telling figure, because keyboards live or die on daily habit. Four million people typing in Sinhala natively from their phones means the input model works well enough that nobody has to think about it.

Published under the name Bhasha, the app’s core idea is simple: give Sinhala speakers a keyboard that behaves like a modern smartphone keyboard instead of a workaround.

Smart Phonetic vs Smart Wijesekara

Helakuru supports two Sinhala layouts, and the choice matters more than it might sound.

Smart Phonetic works the way most bilingual typists think. You type Sinhala sounds using Latin characters and the app resolves them into Sinhala script as you go. It has almost no learning curve for anyone who already types English on a phone.

Smart Wijesekara is the digital form of the traditional Sinhala typewriter layout, the one used by trained typists and people who came up through offices and print. Muscle memory built over years transfers directly, which is exactly why the layout still exists.

Word prediction runs on top of both, and it covers English as well as Sinhala. Voice typing is available for both languages too, which is useful for longer messages where thumb-typing a complex script slows you down.

The practical result is that switching between Sinhala and English mid-sentence, which is how a lot of everyday Sri Lankan messaging reads, stops being a chore.

Helakuru for PC

Phones are where Helakuru was designed to live, and there is no official desktop build. That has not stopped people from wanting it on a bigger screen, usually for the same handful of reasons: writing longer Sinhala documents, handling social media accounts, preparing captions and posts, or working through a backlog of messages with a full-size keyboard in front of them.

The route to that is an Android emulator, and guides that walk through how to download Helakuru Sinhala Keyboard for PC generally follow the same broad shape regardless of which emulator you pick.

Emulator setup

The general process looks like this:

• Install an Android emulator on Windows and let it complete its first boot, which can take several minutes.
• Sign in with a Google account inside the emulator so its app store works normally.
• Search the store for Helakuru and install it the way you would on a phone.
• Open Android settings inside the emulator, enable Helakuru under languages and input, then set it as the active keyboard.
• Open any text field in the emulator to confirm the keyboard appears and the layout you chose is active.

Enabling the keyboard at the system level is the step people miss. Installing the app alone does nothing visible, because Android treats input methods as something you explicitly switch on.

The tradeoffs of running a phone keyboard on Windows

Running a phone keyboard inside a desktop window changes a few things:

• Your physical keyboard usually types straight into the emulator, which can bypass the on-screen Sinhala keyboard entirely unless you disable hardware input for that app.
• Text you produce lives inside the emulator’s Android environment, so moving it to Windows means using the emulator’s clipboard sharing or a shared folder.
• Voice typing depends on the emulator passing your PC microphone through, and that permission has to be granted on both sides.
• Performance is tied to virtualization support being enabled in your BIOS, which is the usual cause of a sluggish emulator.

None of this is exotic, but it explains why an emulated keyboard feels different from a native one. You are running a full Android system to host a single input method.

For someone drafting a long Sinhala post, that tradeoff is often fine. Smart Wijesekara on a real keyboard, with prediction filling in the rest, is close to the experience trained typists already know, and it beats pasting text through a browser converter every few lines.

When people picture a data center, they picture the glow. Rows of blinking servers, neat cabling, the hum of something important. What they almost never picture is the floor underneath it all. And yet the floor is often what decides whether the room stays alive or cooks itself into a very expensive paperweight.

The AI boom has turned up the volume on this question. More power in, more heat out, more pressure on every square foot. So what’s happening beneath the servers that keep your apps, your inbox, and your streaming queue running?

The Room Is a Machine, and the Floor Is Part of It

A data center isn’t a warehouse full of computers. It’s one giant machine with a lot of moving parts, and airflow is one of them.

Heat is the enemy of anything with a circuit board, and servers throw off plenty of it. Pull that heat away efficiently and the equipment lives longer. Fail, and things get expensive fast.

That’s where the raised floor comes in. It’s a modular platform sitting above the structural slab, creating a plenum, a hidden air highway, underneath the room.

Cold air pushes up through perforated tiles right in front of server intakes, and the same hidden space carries cables, power runs, and water lines. The room looks tidy on top because the mess lives below.

The temperature rules aren’t arbitrary either. ASHRAE guidance recommends server inlet temperatures between 18°C and 27°C for reliable operation and reasonable energy use. Hit that window consistently and your gear behaves. Miss it, and you pay twice, once in power and once in shortened hardware life.

Small Temperature Choices, Big Money Consequences

Cooling isn’t a rounding error on the utility bill. At efficient hyperscale sites it can be a modest slice of total consumption, but at less-tuned enterprise facilities it can balloon into a much larger share. That spread is the difference between a well-tuned airflow strategy and one that leaks cold air into empty corners.

The setpoint matters more than most operators assume. Nudging the inlet temperature up a degree or two, when the equipment tolerates it, translates into meaningful cuts in cooling energy. Multiply that across a room running around the clock and the savings show up fast on the P&L.

The catch: you can’t safely raise setpoints if you don’t know what temperature the servers are seeing. In raised-floor rooms, air near the ceiling can run 5 to 10°C warmer than air near the floor. Sensor placement, containment, and tile layout stop being cosmetic and start being financial.

What the AI Wave Is Actually Doing to the Room

The demand curve has bent. Global data center electricity use is projected to roughly double to around 945 TWh by 2030, growing more than four times faster than overall electricity demand. That’s not a rounding error. That’s a new utility.

Rack density is climbing to match. Uptime Institute’s 2024 global survey pegged average typical rack density at 8 kW, with some sites already reporting racks above 50 kW. The same survey put the industry-average PUE at 1.56, a number that has stubbornly refused to move for years.

Here’s the honest tension.

Traditional under-floor air distribution starts to strain once you push into the 8 to 10 kW per rack range and up. That doesn’t kill the raised floor. It changes its job.

In modern builds, the space beneath the tiles is increasingly a service corridor for liquid cooling lines, high-amperage power, and structured cabling, not merely an air duct. Well-designed data center flooring systems are being reengineered around that reality: higher load ratings, better grommet sealing, panels that swap out without dropping a row, etc.

How Smart Operators Are Thinking About the Floor

The mindset shift is subtle but real. Most teams still picture the floor as a passive surface; in practice, it’s an active piece of infrastructure that either helps the room hit its efficiency and reliability numbers or erodes them without anyone noticing.

• Seal the leaks. Every unsealed cable cutout is a slow bleed of conditioned air. Grommets and brush seals pay for themselves in months, not years.
• Match tile to load. Perforated tiles belong in front of active racks, not scattered by habit. The airflow map should match the heat map.
• Plan for change. Racks get denser, workloads move, hardware refreshes. Panels rated for higher loads and easy lift-out save future headaches.
• Measure at the inlet. Ceiling and return-air readings hide problems. Sensors at server intakes tell you what the equipment is breathing.

The Boring Layer That Keeps the Interesting Layer Working

There’s a broader lesson here for anyone running a business, not only a facility. The parts of an operation that never show up in the pitch deck (plumbing, flooring, airflow) often decide whether the flashier stuff works at all. Founders who obsess over the visible surface and skip the substrate tend to pay for it later, usually at the worst moment.

Content strategy works the same way. Marketers who chase the shiny top layer, the campaigns, the launches, without doing the substrate work of niche audience research, tend to see the same flat returns operators see when they neglect airflow. The interesting layer only performs if the boring layer is right.

So the next time you scroll past a slick product demo or a cloud service that magically scales, spare a thought for the floor beneath it. It’s holding up more than servers. It’s holding up the whole illusion that this stuff runs itself.