Blog — Ed Tsarionov

Don't like something? Must be AI

Mon, 07 Sep 2026 00:00:00 GMT

Lately I've noticed a new trend: people see AI everywhere. Well-written text (with em dashes, of course)? Definitely AI. An animation or art style that isn't the usual one? AI. New movie out? Let's find the AI in it...

You've probably seen the posts like "how to tell if a text was written by AI," or the videos of a teacher running an AI detector on student work. (For the record: none of that works)

Being skeptical about what you see online is a useful skill. But AI-hunting is turning into hysteria and a convenient tool for manipulation. Don't like something? Must be AI.

And there's no way to answer a comment like that, because AI can fake almost any proof you might offer — from drafts of the text to animation storyboards. So it becomes an easy, unanswerable way to undermine trust.

I understand why people react this way. Part of the blame is on creators (myself included) who use AI in their work and don't always say so. And on the big companies laying people off and trying to replace them with AI.

But if you're walking around the internet with a magnifying glass looking for AI, ask yourself: are you really doing it to "know the truth"? Or are you just on a witch hunt for the social approval?

Fun little projects, and the shrinking gap between idea and execution

Sun, 06 Sep 2026 00:00:00 GMT

It so happened that I was both editor and scriptwriter for several projects about AI. When they were just starting to appear, we made two seasons of the podcast «И тебя заменят» ("It'll Replace You Too"), where we explained how AI work with references to pop culture. Then there was a special season of «Хакни мозг» ("Brain Hack") about how similarly (or differently) our brain and artificial intelligence work. And this summer I helped a little with the YouTube show «Репортаж на диване ("Couch Report") — where, for script development, I ran vibe coding and agent setup experiments on myself.

So you could say I've been documenting and analyzing the development of AI for the last three or four years. And if you listen to the first episodes of «И тебя заменят» — we make predictions there that have already come true. That's how fast this development is moving.

The main result of this development (for now) is not the tens of thousands of layoffs (although those too), but the shrinking gap between an idea and its execution.

Five years ago, to create, say, a Telegram bot, I needed to find a builder, learn a programming language, or pay someone to make it. Three years ago — I could do it by talking with ChatGPT, but it would have taken about a week and I still would have had to write the code myself. Today I can come up with a project in the morning and implement it by evening with the help of Claude Code.

For example, my friends and I watch English football and discuss it in a chat. Yesterday I thought, oh, it would be fun to run a tournament with match result predictions. A couple of hours — and I had a ready bot that records predictions, tallies scores, builds a standings table, knows the match schedule and can send deadline reminders. A fully fledged game just for the chat.

And these little fun projects are the perfect indicator of that shrinking gap. Spending resources on a game for 5 people used to simply not be worth it. But now anyone can create a personalized tool for friends, family, or themselves. One that entertains, makes life easier, or is just nice to have.

This gap has become so small (and will keep shrinking) that even the main question of digital entrepreneurship has changed. Before, that question was: how do you execute an idea. Whoever executed better, whoever had more resources, won.

Now the question is — what exactly is the idea? And how we will distribute it? Because now anyone can execute — but coming up with an idea that's viral enough, that's another story.

YouTube changed what a view is

Fri, 04 Sep 2026 00:00:00 GMT

YouTube now counts statistics differently for longer videos.

Before, a regular view was shrouded in mystery — some said it required more than a minute of watching, others said seven seconds. But it was clear that it wasn't just any press of the play button. YouTube kept this metric secret.

Now, starting from late August, any instance of playing a video will count as a view (autoplay from the feed seems to be an exception). What used to count as views has now become "engaged views" buried in analytics (still with no explanation of how it's calculated).

As a result, view counts on new videos have grown by 2--3x. And the larger the channel, the bigger the gap between the view counter and engaged views. The latest episode of "Baka!", for example, shows 63k views but only 21k engaged views.

YouTube clearly benefits from this — if TikTok and Instagram can count anything and everything as views, YouTube doesn't want to fall behind. Otherwise their numbers would look less attractive to advertisers and investors.

Potentially this works in creators' favor too — not every advertiser, let alone agency, will bother digging into all of this. Which means ad space can be sold at a higher price.

On the other hand, inflating your view count has become even easier, public counters will ultimately be trusted less — and analyzing channel popularity, trends, and finding great creators will only get harder.

What do you think about this decision?

A YouTube chart for one Baka! episode: 63,720 views against 20,859 engaged views

How I built a custom podcast website in 11 days, with AI and no coding

Tue, 18 Aug 2026 00:00:00 GMT

Part I. Why a website at all

My podcast, Baka!, is seven years old. It started as two beginners talking about anime and grew into a media project: a daily Telegram channel with notes about animation, video essays on YouTube, articles for Kinopoisk and T—J, 120,000 followers on TikTok.

The problem was that all of it lived inside social networks. Which meant it was almost invisible — to search engines and to AI alike. Seven years of conversations about anime existed only inside platforms, each with its own search and its own rules.

A year and a half ago — before AI agents existed — I had already tried to build a site with ChatGPT. I got as far as registering a domain and stopped. The instructions were incomplete. Too much had to be done by hand: copying pieces of code somewhere, working out why the thing that should run didn't. The model gave advice, often out of date, and couldn't do the work for me. I suffered for a couple of days and gave up.

Then, two weeks ago, I saw a post by Samat Galimov, who makes Zapusk Zavtra, one of the best-known tech podcasts in Russian. He had built a site for his show, with transcripts of every episode, over a couple of evenings.

So I decided to try again. The idea was this: if you can make transcripts of a podcast, you can make a full catalogue of the whole project, with cross-references. So that you could click on the name of an anime and see everything we'd ever said about it — in posts, in articles, in podcast episodes, down to the minute in the conversation.

That's where the structure came from: the site is two halves, an archive of material and a catalogue of anime, and what links them is mentions of titles. The address is bakapodcast.com.

The first commit

Part II. How the work worked

The last time I wrote code was in seventh grade, in Pascal. So I started planning the project straight away in Claude. I didn't know what tools it could use, what content management systems existed (apart from WordPress), how a static site differs from a dynamic one, or what the limits were. So I asked it to ask me questions.

After several rounds of questions and my extra explanations of the idea — for example, I had to explain transcripts separately and send the Zapusk Zavtra site as an example — Claude proposed a stack: Astro as the site generator, Sveltia as the admin panel, hosting on Cloudflare Workers.

We rejected WordPress and site builders because there are no ready-made plugins for what I had in mind anyway. I would have had to build the same things, only inside someone else's system.

It's worth saying that at this stage I kept asking Claude stupid questions. Is this free? Why are we taking this particular admin panel? Will there be any limits on the design? Before making a choice, I wanted to understand why we were doing what we were doing. So much so that for the first few days Claude wrote "Don't worry, this is free" in almost every message.

"Don't worry, this is free"

Out of those discussions Claude wrote me eight specification documents (by the end of the project there would be twice as many). It explained how to switch on Claude Code and generated the prompt for the first session.

From then on I used exactly that pair: Claude in the browser and Claude Code. The first explained the work to me, turned my ideas into specs and wrote prompts; the second implemented them and wrote the code. All I did was move text from one window to the other.

A prompt written by Claude for Claude Code

Out of that grew an idea, at once brilliant and completely obvious: I asked Claude Code to write reports about its work into separate documents. So that I could feed them to the browser Claude afterwards, instead of retelling things in my own words.

That paid off many times over, because if I hadn't done it, the project would have fallen apart somewhere in the middle. Claude Code would have run out of context, started lagging and redoing what it had already done. Instead I could calmly start new sessions without fear of losing progress.

The first commit — the skeleton of the site — appeared at 14:20 on 5 August. By the end of that evening there were forty commits: the home page worked, so did writing posts and sections, and the site opened on the internet — without a domain and without a design for now.

The first version of the site

Part III. Three days of design

The clearest way to explain how vibe coding worked for me is the design of the site. In the original spec Claude generated, the design was supposed to appear by itself: I'd send a couple of references and Claude Code would invent the rest from them. It probably would even have worked. But I wanted more control.

To save money (more on that below), I decided to think through the design separately, in ChatGPT, and hand it to Claude Code afterwards. That was a mistake.

I gave ChatGPT a few sets of references: the design the podcast already had; a home-page layout in the style of a magazine; layout and typography; and an idea of stamps for the anime cards. And I asked it to draw a design based on them. ChatGPT read the style I wanted to arrive at quite accurately:

"The main site looks like a modern cultural magazine. A light paper background, plenty of air, thin black lines, large typography, but without excess decoration. The home page uses an editorial grid with materials of different weight rather than standard cards."

But what it could draw was only an approximation of a page. Every generation differed in the details. It couldn't carry my precise corrections into new versions. After several fruitless attempts it suggested working with HTML files instead of generating images. Great, exactly what I need, I thought (the problems didn't end there).

ChatGPT's first mockups

With HTML things went faster. I asked ChatGPT to draw a page, gave it corrections, it redrew — and so on until I was happy with the result. The home page took me eight iterations, the article mockup four. When we had made (as it seemed to me) all the pages, I asked ChatGPT to write a document describing the rules of the design system: which spacing where, fonts, blocks and so on.

Then I went to Claude, handed it ChatGPT's work and asked it to fix any mistakes. Well. Claude simply destroyed ChatGPT's work. Just one example:

Claude reviewing ChatGPT's design system

I spent the next four hours getting Claude to remake all the mockups according to the rules ChatGPT had written for itself and then failed to follow. In the end, on top of those eight iterations of the home page, there were another eighteen builds of the whole site from Claude.

And even then I hedged. Before handing what seemed finished to Claude Code, I asked a separate chat whether I had prepared all the pages needed for development. Guess the right answer: of course not. It turned out I had forgotten the error page, the states of the audio players, what links look like and a dozen other small elements you don't even think about when you look at a site with an untrained eye.

That still wasn't the end. It's one thing to draw mockups and quite another to transfer them correctly into code. The transfer took me the next two days, thirty Claude Code sessions and more than 20% of all the edits in the project. That is: drawing the whole design took an evening. Making it work took two days almost without a break.

Project statistics from Claude

Why so long? An example. The site has a divider, a separating line. Simple, you'd think: one line, the same on every page. In the mockup it's one picture. In code it took six attempts over two days. Each time it turned out some other element also drew that line, one nobody had thought about: the border around the search box, the seek arc on the player, the edge of a pull-quote. One page of text from the original mockup turned into five different ones — with a cover and without, with a transcript and without, with video, on a phone. A mockup shows one state; a site needs all of them. If you don't tell Claude Code what those states should look like, it will invent them itself. And not always the way you need.

That is probably the main rule of vibe coding: you can not know how you want your task done, but you absolutely must understand what the task is. And what the result should look like.

Part IV. The big problems

The real difficulties appeared only towards the end. When the site already worked, looked decent and could be used. Because I started moving in the content and giving it to friends to scroll through.

And it turned out the site didn't open in Russia. It was hosted on Cloudflare, which Claude had picked on day one, and that was a time bomb. Since June 2025, sites that run through Cloudflare have barely opened in Russia: the state internet regulator, Roskomnadzor, restricts not individual sites but the service itself, so everyone who lives there gets hit, regardless of content. For a podcast whose audience is mostly in Russia, that's a site nobody can see.

Not a cheerful conversation

I had to find a way around it, fast. The first idea was to move everything to the hosting I'd bought for embedding video. But it turned out that was blocked in Russia too, by exactly the same logic.

What saved me was that the site is static: Astro doesn't run anything on the server, it builds ready HTML files in advance, which just need to be laid out in folders. A folder like that can be put anywhere — including in two places at once. So I put it in Yandex Object Storage. It can serve static files straight to the browser, it's Russian and therefore not subject to the restrictions, and it plugs into Yandex Webmaster, the Russian equivalent of Google Search Console. The move took another day.

After that, loose ends kept appearing. In a couple of days I used up Yandex's free quota of operations, because I hadn't thought in advance that they needed optimising and Claude Code hadn't thought of it by itself — luckily the cost came to a few rubles. Then all the videos moved there too from the original hosting, because dragging them across two platforms every time turned out to be more expensive than storing them in one place.

"18 rubles already!"

The funniest problem appeared the day before launch. On 13 August I spent the whole day in the admin panel, sorting through the archive of posts moved over from Telegram. The admin panel is built so that every save is a separate record in GitHub: you press "save", and the edit goes straight to the code repository. Over the day I racked up 512 such records. To GitHub that looked not like a person working but like a bot, and its anti-abuse protection flagged my account as spam. The admin panel stopped saving, the site stopped updating, and I was left with half the posts unedited. Panic, sadness, postponing the launch — all of it looked inevitable.

I immediately started asking Claude what to do. Wrote to support (the reply still hasn't come), thought about connecting a different account to the admin panel (the AI talked me out of it), and in the end the workaround turned up by accident. It turned out that although the admin panel couldn't push changes to GitHub, the Claude agent could! So we ran the admin panel on my computer, locally, and it started saving changes not to GitHub but straight to disk. I edited as much as I needed, and then sent everything at once with a single command through Claude Code. My edits went up in one piece: eight thousand lines at a time.

I even had to write this situation down as a separate rule in the project's instructions, because Claude Code kept forgetting about it and publishing the site the usual way, which is how I twice looked at the old version and said nothing was working.

And because of it, right now I have to add posts and podcast episodes by hand. Even though Claude and I designed an almost fully automated system.

Part V. Robots that work on their own

Here they are, left to right:

The nine robots

The site now has nine robots working on it (or rather, they will, once GitHub unblocks me).

The first is the podcast robot. Every morning it checks whether there are new episodes in the RSS feed and, if there are, creates a draft on the site — with a title, a player and a description.

The second is the transcript robot. If there's a new episode, it takes it and runs it through ElevenLabs Scribe. In the transcript it places timecodes and can tell the hosts apart by the pitch of their voices. It puts the result into the same episode draft.

The third is the cover robot. It cuts the character out of the cover of each new episode so they look good on a transparent background. And adds it to the same draft.

The fourth is the Telegram import robot. Every six hours it goes into the podcast's channel and collects new posts. It turns them into drafts: with a title, a cover and text. If a post has several images, it makes them into a gallery. And it automatically replaces links to other Telegram posts with links to the same material on the site.

The fifth is the tagging robot. It automatically checks all new drafts on the site against the anime catalogue. If it sees a match, it puts links to the titles into the posts (and the transcripts) by itself.

The sixth is the new-titles robot. It's run by hand and goes through new posts. Its job is to find mentions of anime that aren't in the catalogue yet. It checks words in quotation marks first, but goes through the rest of the text by its own rules too. If it finds matches, it writes them out on a separate page in the admin panel, where I can mark by hand which of them are anime (Death Note) and which aren't (Shrek).

The seventh is the catalogue robot. It picks up the mentions of new titles and builds pages for them in the catalogue. It takes the information from two sources, Shikimori (a Russian anime database) and AniList, in case one of them gets blocked.

The eighth is the grammar robot. Russian nouns change their endings depending on the case, so a title can appear in a text in half a dozen forms. This robot generates all the forms of each new title so that a mention is found whatever form it's in.

The ninth wakes up once a day and updates the Russian copy of the site.

Besides these, Claude Code and I wrote a few one-off robots. One, for instance, collected my old articles from DTF, a gaming and pop-culture site, and another laid them out by template. Without the robots, going through almost fifteen hundred drafts would have taken me far longer than three days.

So my work on the site now comes down to checking the drafts and mentions the robots have prepared and approving them for publication. Minimum manual labour, maximum automation.

(A reader asked whether the robots are AI agents. They aren't: Claude took part only when they were written; after that everything runs by itself. Some of them hang on Cloudflare cron triggers — scheduled publishing, the RSS check, updating the Russian mirror. The rest run on cron on a small Hetzner server: everything slow and heavy, like transcription through the ElevenLabs API, cutting the character out of the cover, the Telegram import. Of all nine, only two call an AI: the transcript robot and the cover robot. The others are ordinary code — parsing RSS, requests to Shikimori and AniList, finding titles in text against a dictionary, generating word forms. To save tokens and not run to the neural network more than necessary.)

Part VI. And how much does it cost?

Far less than a programmer's work — not to mention the designer, the front-end developer, the editor and the two or three other specialists you'd have needed for a job like this before.

The Claude subscription is $20 a month. I raised the plan to $100 as a one-off. If I'd stretched the work over a month or two, I'd easily have fitted into the basic price.

The ElevenLabs subscription for transcripts is $22 a month. That was enough even for the insane volume of 143 archive episodes (140,947 lines) without paying extra.

The domain, bakapodcast.com — about €10 a year, bought on porkbun.com.

Yandex Cloud — about 20 cents, for exceeding the limit of 10,000 operations a month. Most likely it will come to under €3 a month even with thousands of visitors. I may need to buy more storage later, but our videos mostly go out on YouTube and the audio lives on Mave, our podcast host, so the free tier covers the images for now.

Unnecessary spending: Hetzner hosting, €9 a month.

Astro, Sveltia, Cloudflare, GitHub — free.

Total: building the site cost roughly €130, and running it costs about €40 a month (or around €20 if you unsubscribe from Claude).

Part VII. The result

The site came together in eleven days. That's 87 Claude Code sessions, 628 messages from me in the terminal and about a hundred hours at the keyboard. The site now holds 143 podcast episodes with full transcripts, almost a thousand posts and a catalogue of 507 titles with 14,286 mentions across the material. The thing it was all for works: you click on the name of an anime and see everything we've said about it — in notes, in articles and in episodes, down to the minute in the conversation.

It lives at bakapodcast.com, and for Russia there's a copy at ru.bakapodcast.com.

Having finished this version, I'm already thinking about what to add next: social features — comments, ratings, achievements for titles watched and episodes heard. And I'm absolutely sure it's within my reach.

By the way, in the podcast trailer that opens this text, the only thing not made by AI is the sound of the keyboard. Everything else — the music, the animation and, as you now understand, the site itself — is Claude's work.

Special projects are coming back

Mon, 17 Aug 2026 00:00:00 GMT

When I was in school, one of the most exciting things in media was a format we called "special projects": large interactive longreads that combined text with video, audio, and infographics.

I still remember the thrill of reading "Snow Fall: The Avalanche at Tunnel Creek," the New York Times piece that won a Pulitzer in 2013.

Back then, only major outlets could afford the format. It took more than journalism — coding a site from scratch, building graphics and animations, designing charts and diagrams.

That's no longer true. Ilya Inozemtsev single-handedly produces large research pieces with custom-built websites: about jokes, or, most recently, about rappers' vocabularies. And I built a full website for my podcast on my own, in a little over a week.

Vibe coding opens the door to every kind of interactive and multimedia project. It's especially exciting for niche ideas nobody would have touched before, simply because they'd never pay off.

Add the fatigue with feeds and algorithms, add the steady flow of good journalists leaving legacy media to go solo — and I expect the next few years to bring a new dawn for special projects

The therapy of failure

Sun, 16 Aug 2026 00:00:00 GMT

Once I was taking a taxi to football training. I was already in my boots and kit, because there was nowhere to change. The driver noticed and got curious. Where did I play, who with, how seriously I took it. And then he asked: "So, are you any good?"

I thought for a moment and answered honestly: "Nah, not really."

For the rest of the ride he chuckled at my answer with great satisfaction. He was delighted that I could admit so honestly and calmly that I wasn't good at something. And that I didn't stress about it.

I often think about that conversation and that reaction.

Ever since school, we've been taught that not knowing something or not being able to do something is bad. On your CV you'd better add a couple of years of experience, and your English had better be C1. On social media you should only post smart thoughts and photos from expensive vacations. And in general, in everything you do, you must appear (or actually be) the best. I'm prone to that kind of thinking myself.

But in reality — on a human level — people are quite warmly disposed toward those who admit their shortcomings. There's no fun in kicking someone who's already down. There's even the Benjamin Franklin effect, which says that someone who has done you a favour will feel more warmly toward you than someone you've helped yourself. So asking for help is neither scary nor bad.

And doing something you're not good at is actually a form of Alzheimer's prevention. And of pride prevention, too.

I still play football pretty badly, but I go twice a week anyway. Every time, it's a humbling experience. Can recommend.

Three quick facts about vibe coding

Tue, 11 Aug 2026 00:00:00 GMT

I have three quick facts about vibe coding for you:

It works (just in case you still thought it was a myth)
It's addictive (you literally get hooked on the feeling of excitement from bringing ideas to life so fast)
It takes time and effort (it's not a magic button that does everything instantly and effortlessly)

What if it's actually easy?

Thu, 16 Jul 2026 00:00:00 GMT

Call the dentist and make an appointment. Prepare the documents for a visa. Cancel a subscription you no longer need. Congratulate a colleague on their birthday. Return clothes that didn't fit.

Just writing these sentences makes me break out in a cold sweat. I hate these kind of organizational chores. Things that sit somewhere between work and leisure. Bureaucracy in the broadest sense.

Unfortunately, life (and especially life as an immigrant) seems to mostly consist of exactly these kinds of tasks. So I'm always looking for ways to stop being afraid of them.

A few days ago, I came across an idea I want to try. Oliver Burkeman, author of The Imperfectionist, suggests asking yourself a simple question before starting a task:

What if this is actually easy?

A large part of the anxiety these chores create comes from uncertainty. What if the visa is rejected? What if I don't know what to say on the phone, or which language to use? What if they refuse to exchange the clothes?

But asking yourself "what if it's easy?" shifts your attention to specifics

"...to the actions you can actually take. The point isn't to imagine some parallel universe where the task would be effortless. It's simply to allow yourself to consider that it might, in fact, be easy in this universe.

I realize this sounds like the worst kind of denial — as if you could get what you want by commanding the universe to align with your wishes. But in practice, it can work surprisingly well. Not because it's some mystical order to the cosmos, but because it reminds you not to fall back into the old habit of making things more complicated than they need to be, or adding a sense of painful effort where none is actually required."

Or, as one of my classmates used to answer when people asked how she was the only person in class who could solve difficult math problems:

"You just sit down and do it."

There is no such thing as a bullshit job

Thu, 09 Jul 2026 00:00:00 GMT

Seven years ago, I read David Graeber's Bullshit Jobs and found the whole concept deeply annoying. Today, it seems even more harmful to me.

Graeber defines a bullshit job as:

A form of paid employment so completely pointless, unnecessary, or harmful that even the employee cannot justify its existence — and yet, because of the conditions of employment, they have to pretend otherwise.

Graeber puts designers, bloggers, lawyers, and all kinds of other people into this category of supposedly useless workers.

But here's the problem: the value of a job is not primarily determined by the worker. It is determined by the employer. If the employer believes that the work being done — or even not being done — justifies the cost of the salary, then the job has some kind of value. That value may come from personal ambition, altruism, badly designed processes, or something else entirely. But it is still there.

Just because something seems useless to you does not mean it actually is.

So why is this idea even more harmful today? Because of artificial intelligence.

AI can quickly and efficiently generate images — so designers are no longer needed. It can write code — so programmers are no longer needed. It can give medical advice — so doctors are no longer needed. And so on. If we accept Graeber's formula and factor in AI, we end up with the conclusion that almost all human work is already bullshit.

And, honestly, that sounds pretty doubtful.

That is why we should not define the value of work only through its final result, or through some hypothetical usefulness. In fact, we often instinctively resist this framework, because we sense that our own work, and other people's work, contains something beyond the simple execution of functions.

There is work I dislike, and it is probably better to avoid it when possible (and Graeber rightfully points it out). There is harmful work, at least in my view. There is mechanical work, which robots will eventually do — and thank God for that.

But there is no such thing as a bullshit job.

Why counting podcasts no longer makes sense

Wed, 20 May 2026 00:00:00 GMT

Five years ago, I did a research project trying to estimate how many podcasts were being launched in Russian.

In English-speaking markets, there were already aggregators like Podcast Index constantly tracking the number of new shows. But for the Russian-language market, there was almost no reliable data. So it felt important — a way to understand the actual state of the industry.

The logic seemed straightforward: if lots of new projects keep appearing, podcasting is growing. If not, it's stagnating.

But today, this metric — in Russian, English, or any other language — no longer tells us much about the health of the industry.

Because roughly 40% of all newly launched podcasts are now AI-generated. At least in English-speaking markets for now, but the same trend will inevitably spread elsewhere.

There are entire companies built around mass-producing AI content. Inception Point AI, for example, reportedly operates tens of thousands of active shows.

The problem with this wave of AI slop — beyond the obvious quality concerns — is that it no longer correlates with the number of actual creators, audience demand, or cultural interest in podcasting itself.

It's essentially a speculative bubble: content generated at scale in the hope of capturing even a tiny slice of audience attention.

Most of these podcasts have extremely low ratings and engagement (though there are exceptions). So I'm not ready to declare that AI will replace podcasters anytime soon. But it already makes many of our traditional ways of measuring the market far less meaningful than they used to be.

What is a podcast?

Mon, 26 Jan 2026 00:00:00 GMT

Debates about what a "podcast" actually is have been going on since the dawn of time. I have no proof, but I wouldn't be surprised if Roman legionaries were already arguing about whether something that doesn't come in an RSS feed can really be called a podcast (if you have no idea what I'm talking about, you're a lucky person).

Today, these debates have completely hit a dead end. Because "podcasts" now include video interviews on YouTube that never were—and never will be—on podcast platforms. And fake TikTok videos with a specific visual aesthetic. And voice messages people publish in Telegram channels.

The question "what is a podcast?" has shifted from a practical one to a philosophical one. Steve Goldstein, founder of Amplifi Media, calls this the Liquid Era of podcasting.

The essence of this era is that an audio-only product is no longer enough (I'd even say this became obvious in the Russian-speaking space a couple of years ago). The linear approach no longer works: record an episode, post an announcement on social media, get listens.

For a "podcast" to live—let alone make money—it has to be everywhere: in short-form videos, on YouTube, in Telegram channels, on Twitter. And it can't all be the same announcements reposted everywhere. It has to be different formats for different platforms: quotes, clips, visuals, landing pages, articles.

The «Бака!» podcast has two Telegram channels, a chat, TikTok, YouTube, a DTF channel, a VK group, Threads, and Instagram. And almost everywhere, we publish adapted content.

For each episode of «Дар места», we release a photo album from the locations, and by the end of the first season we even published a small-run print magazine. Alongside «Колыма», we launched a stunning website with long reads and character cards.

A "podcast" today is no longer just recording audio into a good microphone and cutting tracks in Audition. It's also about understanding how to repackage that content into a dozen different formats (and you can safely stop worrying about what an RSS feed is).

A "podcast alignment chart": nine definitions of a podcast, from "an audio RSS feed" to "anything on the internet"

The unanswered questions

Wed, 11 Jun 2025 00:00:00 GMT

Storytelling, especially online, is all about asking the right questions.

A good topic, a nice thumbnail, a catchy headline — that's usually enough to spark interest. But it's not enough to keep someone engaged all the way through.

To keep them engaged, you need to give them a puzzle to solve:

• Can the blogger tell Coke from Pepsi?

• Is it worth buying the new iPhone?

• Why did Manchester lose again?

• Is life really meaningless?

• And when will the dollar finally crash?

(Of course, everyone has their own set of questions.)

Once the question is asked, once there's a reason to keep reading or watching, you guide your audience toward the answer. Along the way, you provide context, raise more questions, and sometimes even play with their expectations.

Without that tension, content can still be beautiful and useful — but it won't work as a story.

It's not about quality, it's about recognition

Sat, 17 May 2025 00:00:00 GMT

Last week, I was on the jury of yet another podcast pitch event — and I noticed something interesting. Every single student used AI to generate their cover art. That fact alone speaks volumes.

But what really caught my attention was something else. I realized that the covers that looked obviously AI-generated — even without weird artifacts or mistakes — put me off. Meanwhile, those that managed to hide their artificial origin felt more polished and appealing. Why is that?

The better AI gets at generating images, the more I realize: the issue isn't about quality — it's about recognition.

You look at an image, and suddenly it clicks: "Ah, this was made by AI." And just like that, its value drops. It stops feeling alive. It's as if your brain has a built-in trigger for valuing human effort — or noticing its absence.

But if that recognition doesn't happen, the AI image retains its value — even if it's technically bad. That's why low-effort AI slop still thrives on platforms like Facebook — the audience there simply doesn't notice it's artificial.

Which leads me to a bigger question: what if this mental mechanism isn't limited to AI? What if we tend to devalue any work that feels like it "just happened"? But once we associate it with a specific person — it suddenly feels alive, cared for, meaningful?

If that's the case, maybe we should train ourselves not to detect AI-generated content (since soon we won't be able to), but to notice and appreciate the invisible human work all around us.

The tyranny of purpose

Wed, 14 May 2025 00:00:00 GMT

Yesterday, Alexey Tkachuk released a video where he reflects on workaholism and the FOMO effect.

In short, the idea is that freelancers (especially those with their own projects) constantly feel like they're wasting time when they're not working. Because they create their own tasks and goals, manage their own time — and as a result, they have no "free time".

I know this feeling very well. I'm a freelancer with tons of personal projects, and my schedule is 90% dependent on my motivation. And procrastination doesn't really fit into this system.

I've come up with a bunch of psychological tricks to deal with this. For example, when I watch anime, I tell myself it's for work — gathering content for my podcast. Or when I play console games, it's to write a review post for my personal channel. Not just because I'm having fun.

I have to trick my brain and create a purpose or meaning for any kind of rest (beyond resting itself). Even though rationally, I understand that the emotions from a great video game are no worse than the emotions from a successful project.

Neuroscientist Anne-Laure Le Cunff calls this "the tyranny of purpose" — the oppressive idea that the activities we fill our days with must lead to some final result or meaning. Because we want to be sure we're not "wasting time."

But in reality, this is just an illusion of control. We have no control over the amount of time we're given or the outcomes of our efforts. The only thing we can control is how we feel — about a day, a year, or life itself.

In the end, the tyranny of purpose leads to nothing but meaningless anxiety (just like FOMO and workaholism, no matter how you name it).

So, we need to learn to do things more often not for some grand result, but simply because... why not?

If you're starting with a blank page, you're already doing it wrong

Thu, 03 Apr 2025 00:00:00 GMT

I've seen plenty of advice on "how to overcome the fear of the blank page" (and many of those tips are genuinely helpful).

But I think there's a more radical approach.

If you're starting with a blank page — you're already doing it wrong. And it doesn't matter whether you're writing a short story, a research paper, or a LinkedIn post.

Starting with a blank page means you haven't done the prep work. You haven't spent time researching, gathering materials, outlining, or making notes.

Because starting to write — isn't the beginning of the process. It's the middle.

Trying to create words and sentences out of thin air, staring at an empty page, is hard and scary. But structuring thoughts, facts, quotes, and ideas you've been collecting and mulling over for weeks? That's much easier.

That's how the "blank page" problem gets solved — before it even shows up.

Content that sticks

Mon, 03 Mar 2025 00:00:00 GMT

The other day, I was going through Patreon's State of Create report on the creative industry, and one particular insight stood out to me.

Here's the data: only 29% of respondents are willing to pay for short-form content, while 49% are open to paying for long-form content.

At the same time, if you ask people what they usually watch, the numbers flip: 61% primarily consume short videos, while only 32% engage with long-form content more often.

Why is that? Why can podcasters successfully monetize their content through community support, while TikTok creators—with vastly larger audiences—struggle to do the same?

I think the answer lies in the difference between content that entertains and content that sticks.

Long-form content simply has a better chance of staying with the audience because people spend more time with it. And we're more willing to pay for what feels important—for what we remember.

Ironically, podcasts benefit from their lack of algorithmic distribution (or rather, their complete absence of it). The more responsibility you have as a viewer to choose what to consume, the more attention you give it—and the more attached you become.

There's a huge difference between an endless TikTok feed that decides what to show you and a podcast app's homepage, where you have to make a choice. And once you do, you stick with that choice for hours.

That's why a few thousand devoted podcast fans can be a far better financial investment than a hundred thousand distracted TikTok followers.

Memorable content > Entertaining content

Patreon survey: 29% would pay for short-form content, 49% for long-form

Stable part-time

Wed, 08 Jan 2025 00:00:00 GMT

I've never worked a traditional full-time job. By chance, I ended up in media with my (so-called) side projects, and there's always been more than one running simultaneously.

Even when I (sort of) settled into a full-time role—first one, then a second, and eventually a third—my side projects didn't disappear.

That's why I always find it tricky when people ask me where I work, what I do, how much I earn, or when my vacation ends. The answer always depends on who's asking.

But that's also the most interesting part for me. I enjoy doing multiple different things, talking and writing about various topics, and developing a wide range of skills.

It would be great if the world had more stable part-time roles—not just in creative industries, but across the board. It would be amazing if there were jobs for three days a week instead of six.

This could open up new opportunities for a diverse range of people: from parents who don't want to spend every moment with their kids to senior executives for whom running an side business could become the norm.

Views no longer follow followers

Fri, 18 Oct 2024 00:00:00 GMT

I came across a very telling chart this morning from the influencer marketing agency Kyra. They tried to track the correlation between average views and the number of followers a creator has. And they found that there is no clear correlation.

Creators whose posts average 100K views can have anywhere from 26K to 1.4M followers.

This is largely influenced by recommendation algorithms — whether something goes viral or not is no longer solely dependent on follower count, but more on luck.

Even in platforms like Telegram, where there are no recommendations, this correlation is becoming increasingly blurred. For example, I've noticed that my anime channel "Baka!" (6K followers) easily outperforms channels with five or six times more followers in terms of average views and engagement.

Most likely, this trend will lead to the conclusion that it's not views that matter, but the communities creators build. These communities will become the only predictable and result-driven source (clicks, subscriptions, purchases), and the focus will shift to paying for access to them.

A chart by Kyra: average views plotted against follower count, with no visible correlation