Computer bytes and human bits ⊙ Buy me a coffee ☕️

Why Short Videos Are Winning

People started reading less because they started reading more.

In 2009, researchers estimated that the average American consumed ~34 GB (≈100,500 words) of information per day. It's hard to argue that the amount of information surrounding us, the amount of text we read across all our screens, has done anything but increase since then.

So yes, people stopped reading for pleasure because the amount of text they read every day exploded.

Reading a book, especially fiction, requires attention and deep context switching. You’re not just reading words - you’re switching into an entirely different mental environment and staying there. Finding time to read isn’t enough. You need to disconnect from everything else and remain inside that context. That’s difficult when the same device you’re reading on is constantly trying to pull you into another one.

This is also my theory for the rise of algorithmic social media - and its latest invention, short videos. These formats don’t require deep attention or context switching. And that’s why the videos are short - the longer the video, the more attention it requires.

You can open them instantly and get a quick dopamine hit delivered straight to the palm of your hand. It’s fast, effortless, and asks nothing from you. And when your breaks are short, you have a few minutes to kill, or you're just tired - it's easier to fire them up than to open a book.

It feels like our modern lives - patchy, quick, fragmented - are designed for exactly this kind of time spending. It’s almost as if we once lived at normal speed, and now technology has pressed Fast Forward.

And if you want to break that cycle, you have to hit Pause first.

Where Is Your Computer?

The future of tech is diverging in front of our eyes.

One camp, the believers in the internet computer, thinks the future belongs to the browser. After all, it’s the logical endpoint of the cloud era. In many ways, we’re already there.

Since this is tech in 2026, they’re AI-pilled. But their real conviction is that the browser is the next OS. AI agents will do everything for you, and it will all happen online, inside your browser window. They believe in it so much that they’re ready to sacrifice everything at that altar, including the Arc browser, which was loved (and is missed) by the community.

The second camp is just as AI-pilled. But instead of the browser, they preach a return to local files and desktop apps, written by AI and running outside it. In this world, software becomes cheap and disposable. If you need a tool, AI just generates it for you. AI agents will still do everything for you, but they’ll run on your computer, working directly with your files. Glaze by Raycast is the latest member of this camp.

Both futures are being built at the same time, so choose ~~your fighter~~ where your personal computer actually is:

in a browser tab,
or on the desk in front of you.

25 years of iPod brain

The article by Molly Mary O'Brien hits hard, especially this line:

It’s hard to believe that there was once a time when consumer technology solved problems we actually had.

I mentioned iPod and dedicated audio players twice recently (1 and 2), and I’m still ready to die on that hill. I genuinely believe Apple made a mistake when it gave up on the iPod and effectively handed the entire audio player market to Android and Chinese manufacturers on a silver platter. Right now, there’s basically no one else in that space.

Yes, the market might be small, limited mostly to niche enthusiasts. Yes, the margins might even be negative. But I’d gladly buy a modern iPod. Not the Touch one, the real one, with a click wheel. It was a fantastic piece of tech.

Why can't one of the most valuable companies in the world have a niche product for enthusiasts? It feels more and more like the iPod was only possible because Steve Jobs loved music and actually listened to it.

In a speech to Apple employees (quoted in Make Something Wonderful, available at https://book.stevejobsarchive.com/), Steve says:

You know, one of the reasons we started doing this [was] we could see that we were getting better and better at iPods, and we could see that there was an opportunity to maybe do the next thing—and what should it be?

And it wasn’t driven by a bunch of market research or financial spreadsheets about how big certain markets were. It wasn’t driven by that at all. It was driven by the fact that we all hated our phones. We talked to all of our friends and all the people we knew, and they all hated their phones.

Ironically, everybody seems to hate phones now too, although for very different reasons.

So maybe, just maybe, history will loop back on itself, and the next big thing will be the one we’ve already had.

In the Age of AI, Human Support Is a Moat

The latest trend is to declare software, and SaaS in particular, dead. Why pay for tools when “AI agents” can build software tailored uniquely to your needs? MacStories calls it the future of software, and Reddit is full of posts along the lines of: “I’m tired of taking notes and forgetting about them, so I vibe-coded a new note-taking app to fix it.”

And in this new era of automated, agent-built software, you don’t even need human support anymore. If AI agents can build the product, surely the same agents can support it as well, right?

The great PM-driven replacement of human support started long before the current wave of AI. Remember chatbots? You’d think they would be much better by now, but nearly every AI support agent I encounter on websites is still awful. At least the previous generation of chatbots relied on deterministic logic, and there was usually some way to reach a human. The new wave of “AI support agents,” however, pretend to be human in the first place, and can hallucinate features out of thin air, features that have never existed. Which, in a sense, makes things worse.

So, to answer the question posed above, I’d say yes, it’s true. You don’t need human support to help customers navigate your vibe-coded software.

But I’d also argue that genuinely great customer support, provided by real humans, is one of the strongest ways to stand out right now and build user loyalty. People are social creatures; they don’t like talking to a wall - even if that wall pretends to be human.

In the age of AI, the companies that stay human will be the ones people stay with.

AI and Intellectual Imperialism

I first noticed it with IT specialists and programmers in Russia.

Their success in one field - and the high salaries that came with it - made them believe they knew better than everyone else. Soon, they started teaching others how to do things in completely unrelated areas, often ones they knew little or nothing about. It’s almost like the Dunning–Kruger effect, but not quite: these people are genuinely competent in their own field; they just overgeneralize that competence to other domains.

Now this intellectual imperialism is spilling everywhere, amplified by AI.

Programming and writing code stand apart when it comes to AI. They have none of the copyright and ownership issues found elsewhere, and nowhere else is its usefulness so obvious, adoption so widespread, or the stigma around using it so minimal. In most other fields, AI is still treated with caution or defensiveness; in programming, it is simply another tool. Perhaps that normalization is precisely what creates the illusion that similar gains must exist everywhere else.

All those AI bros have found a hammer that works great for their nails, and now they’re trying to apply it to every other use case imaginable - even the ones that require a more nuanced approach.

It all reminds me of the square hole meme, where every block is forced into the same square hole. The square hole is generative AI and LLMs, and the AI bros keep trying to fit everything into it. Even if it technically works the first time, that doesn’t mean it’s the right hole for other blocks. The same goes for AI and LLMs: they are not optimal for everything, and the chat interface is not always the best way to interact with computers.

And in this analogy, we’re all Alison on the left - watching, increasingly frustrated, as everything goes into the square hole because those guys “know better.”

Local-First GDPR Governance with Obsidian

This is a nerdy one.

As someone who spends a lot of time reading compliance documents and thinking about GDPR, I decided to see what would happen if I treated governance like a knowledge management problem.

Is a more flexible, lightweight world outside of spreadsheets even possible?

There are plenty of governance solutions out there. Big players like Vanta and Drata are great and all, but a) they cost a lot of money, and b) they’re mostly focused on automation, broader frameworks, and c) they're not really GDPR-specific. They’re also US-based, which means adding yet another processor outside of the EU. And who knows what happens to the Data Privacy Framework in the future.

At the same time, local files are back in fashion. After thinking a lot about Obsidian following my last two posts, I had an idea: what if you could maintain your GDPR compliance documentation entirely in local .md files?

It sounds like the dream of a very particular kind of nerd (me). But could it actually survive real compliance challenges - and the inevitable clash with legal teams who live in Microsoft files? I don’t know, but that didn’t stop me.

So I built a proof-of-concept vault in Obsidian with that goal in mind. This is an exploration of what's possible there right now, and for security reasons I used Restricted mode and core plugins only.

Vault template

Download the compliance vault or clone it from the Github repo.
Unzip the .zip file to a folder of your choosing.
In Obsidian open the folder as a vault.

It’s not as deep as the enterprise solutions mentioned above. There are no mapped controls, no automations, no pie charts or completion dashboards. But the local, interconnected files with backlinks - and the new Obsidian Bases - work surprisingly well for this particular case (take the RoPA base, for example, which pulls in the properties from individual Business Function files). And, the bases also can be exported to .CSV files for audit purposes, when needed.

There’s also a lot more that can be done: local evidence files with structured properties, AI agents with access to local files, and probably a few other things I haven’t even explored yet.

This is a cool little project, and I think it might actually work - if you have enough background in governance and GDPR.

Compliance ≠ complicated tooling.

Obsidian

After my previous post about software, Harrison got in touch to tell me about how he uses Obsidian:

I like using the daily notes plugin to write about what I do every day. There may not always be something significant, but I think it'll be neat to be able to look back on these notes years from now. Plus, everything is stored locally in Markdown files, so it's guaranteed to be future-proof.

And I totally understand why Obsidian is mentioned first. It is a great piece of software, and I really like Kepano’s - now the CEO of Obsidian - philosophy behind it, especially the idea of “file over app.” I use Obsidian myself, in a way very similar to Harrison, although I’m not always consistent enough to keep daily notes.

I’ve also always liked the metaphor behind its name and logo. You start with a raw shard of volcanic glass and slowly shape it into something that fits you alone - a bladelet made by your own hands. There’s a quiet beauty in that idea.

But in our age when people outsource their opinions and decision-making to Reddit and AI, that blade(let) is double-edged.

Obsidian’s greatest strength - its customizability - is also its biggest flaw. It's a classic case of overchoice or analysis paralysis: when there are too many options (plugins, themes, workflows, endless ways to tweak and refine), you end up spending more time sharpening the blade than actually using it.

That’s why I’m drawn to opinionated software. When you find something made by someone else that truly fits, it fades into your daily routine and quietly lets you get on with your life.

Right now, more and more people are building their own tools with the help of AI agents, chasing the idea of something perfectly tailored to them. But I have a feeling that journey has no real destination.

Beyond Reviews

Software ate the world and left no crumbs. Yet it has never been harder to find something worth bringing into your daily life.

Support inboxes naturally skew toward the negative. When software works well, people move on with their day; they rarely write just to say thank you. That’s why, at every company I’ve worked at, we kept a dedicated space for sharing kind words from customers - a quiet source of motivation for the support team and beyond.

I see the same thing happening with writing about software online.

And I get it. When something fits naturally into your day, you don’t feel the need to stop and write about it. You just use it. It fades into the background, like a good chair or a familiar keyboard.

But that’s also the saddest part, because those quiet, everyday experiences are often the most interesting ones. Not polished reviews or launch-day impressions, but the small ways people shape their routines around software that helps them think, read, write, or simply stay organized. That’s where software becomes personal, and that’s the perspective we rarely hear.

Instead, most writing about software today comes from people whose job is to review it. And that creates a strange gap: we hear plenty about features, pricing, and updates, but much less about how these apps actually live alongside someone’s day-to-day life.

It’ll probably get worse, and in many ways it already has. Instead of writing about a piece of software and sharing real experiences online, people are vibecoding their own versions - and then asking an AI agent to post about it on Reddit, Moltbook, or wherever the conversation happens that week.

Is this the future of software? Yes. No? Maybe. Who knows.

I’m doing this the old way. If you’re not a productivity blogger or a reviewer, and there’s a piece of software you quietly enjoy - even just a small detail or a tiny habit you’ve built around it - hit reply below and tell me about it. I'm curious.

I’ll go first.

Following and learning from people you’re genuinely interested in has always worked well for me. Fifteen years ago, I did it by building a carefully curated follow list on Twitter, and it worked great. These days, I’m doing the same thing again - just using RSS and NetNewsWire.

It feels calm and unhurried: just you, a cup of coffee, and a chronological stream of posts from people you chose to follow on purpose. Thanks to Brent Simmons, NetNewsWire is completely free, so it’s easy to start the same way.

Cracks in the Walled Garden

Apple devices are great, but the arrival of OS 26 has taken some of the shine off, and the App Store ecosystem has been stagnating for quite some time already. At this point, I avoid subscriptions that force App Store billing. If your app allows payments outside it, I’m far more inclined to try it.

The abundance of apps on our devices has lulled us into forgetting that this system isn’t open, and that Apple ultimately controls what can and cannot exist on your personal computer (pocket or not), often on a whim.

Want to download an app in your region? Too bad, it’s not available.

That music you paid for? Gone.

Want to install software outside the App Store, like on an actual computer? Nope.

Want to download an opposition app from our glorified App Store? Don’t bother. Your government asked us to remove it, and Apple wants to keep doing business there - so you can fuck right off.

Years before Apple removed ICEBlock from the App Store, it had already been doing the same with Russian VPNs and opposition apps (and I'm sure we can find similar examples in China). What’s changed isn’t the behavior, but the geography: the U.S. is now adopting the same autocratic tactics Russia used years ago, and only now does anyone seem to care.

Cyberpunk and sci-fi taught us to expect a world where corporations are bigger than states - Tyrell, Wallace, Weyland-Yutani. But that world isn’t here. Our world’s corporations may be powerful, but they still fold when states apply pressure, as the Trump administration demonstrated in the U.S.

That’s why I don’t want to give in, or hand everything over to them, or move my entire life into “the cloud.” I still care about having things I actually own, systems I can control, and data that isn’t rented back to me.

The idea of the open web, of freedom, and of ownership is still there, like a lighthouse standing somewhere between a fading past and an unknown future.

It just so happens that Apple chooses to ignore it right now.

Against Becoming a Format

It’s sad to see how much of the modern internet has slid into personal branding.

Once someone stumbles onto a viral format, repetition becomes mandatory -- not because there’s more to explore, but because that’s what the algorithm rewards. And what's even worse, everyone expects it.

The short-video format dominating today’s internet amplifies this effect, producing endless blogs and accounts squeezing the same topic, the same joke, the same idea dry, again and again. Deviating from the formula becomes risky. Try something new, and engagement drops. Change tone, and people ask what’s “wrong.” The internet and the algorithms train people to stay small inside the box that once made them visible. And it's so incredibly boring.

Compare it to Walt Whitman:

I am large, I contain multitudes.

I don’t want to be about one thing. I never wanted to be a one-trick pony -- or a pony at all. I want to be human. A living, breathing human being, with changing interests, contradictions, and new things still left to try.

I don’t want to be smaller to be visible.

Apple Music Mini Player

Apple has created the ideal location for the mini music player on macOS:

Uploaded image — lower right corner is free

But then, they went ahead and made the Mini Player in Apple Music look like this:

Crazy stuff.

The End of “AI Content”

Instagram will soon be so overrun by AI slop that it won’t make sense to label it as such anymore. In the words of Adam Mosseri:

Social media platforms are going to come under increasing pressure to identify and label AI-generated content as such. All the major platforms will do good work identifying AI content, but they will get worse at it over time as AI gets better at imitating reality. There is already a growing number of people who believe, as I do, that it will be more practical to fingerprint real media than fake media. Camera manufacturers could cryptographically sign images at capture, creating a chain of custody.

I explored a similar idea in this blog -- Braincrafted or Heartmade -- though my reasons were entirely different.

NetNewsWire is Moving from Slack to Discourse

NetNewsWire is saying goodbye to Slack and bringing its community to Discourse:

Slack’s been pretty great for us, but it does have some limitations: conversations are automatically deleted and they’re not findable on the web in the first place.

I've been saying the same thing for a while already (1, 2). Closed chat communities are bad, and I will stand my ground:

Messages disappear into the scroll, and good answers get buried in the side threads. It’s like a never-ending group chat - great for real-time energy, but terrible for knowledge that needs to stick around.

That’s why forums still matter.

Forums and comments create structure. They’re searchable, linkable, and persistent. A thoughtful post from three years ago can still help someone today. A conversation can grow over time. And when people contribute there, they’re not just answering a question - they’re building something others can return to.

No Thanks to Public Accounts

I noticed Cooked.wiki in Olly's App Defaults and decided to give it a try, but quickly realized that I won’t be using the service.

Don’t get me wrong -- it looks like a good tool for recipes, and I’ve been cooking quite a lot recently, much more than before -- but there’s one thing that rubbed me the wrong way.

It feels really weird to me to paywall the privacy of your account. Not privacy in a data protection and regulatory sense, but in the sense that your account is public by default, and there’s no way to change that setting without paying.

One could argue that recipes aren’t personal information and that a throwaway nickname would suffice, but it feels like a strange decision.

They’re committed to never running ads, and I respect that, yet this particular choice is… strange.

Imagine if Google didn’t show you ads, but your search activity was public instead. Or if your email provider were free and ad-free, but your emails were publicly visible. Would you use it then?

Maybe it’s just me. It probably is.

If It Quacks Like a Duck...

I can’t see how this could have happened without AI:

How do you paste another company’s email address into your reply if you’re an actual human composing that response? And if it’s a canned answer, the correct email address would already be pre-saved.

The reply from Obsidian didn’t make things any clearer and failed to refute the AI allegations, so I’ll apply a familiar maxim:

If it looks like AI, swims like AI, and quacks like AI, then it probably is AI.