⏱️ Lectura: 10 min
Every time you open YouTube, Spotify, or LinkedIn, someone else decides what you’re going to watch for the next few minutes, not you. That’s the premise of the essay “Attention is all you have”, published on September 21, 2026 on the blog alicegg.tech, which opens with the Tetris effect: play long enough and you’ll start seeing game pieces in clouds, buildings, and everyday objects.
📑 En este artículo
The central argument is straightforward: what you focus on for long enough ends up shaping your thoughts, and less and less of that attention is something you actually choose. The alternative the essay proposes isn’t abandoning the internet, but going back to an older browsing model: bookmarks, hand-picked sites, and RSS instead of algorithmic timelines.
TL;DR
- The essay “Attention is all you have” (alicegg.tech, 09/21/2026) uses the Tetris effect to explain algorithmic attention hijacking.
- YouTube, according to the essay, mixes cooking clips with content about the stock market and war to keep you scrolling longer.
- Spotify inserts “AI slop” between real songs to avoid paying royalties to artists.
- LinkedIn prioritizes opinions from strangers over career news from your actual contacts.
- Before recommendation algorithms, browsing relied on bookmarks and hand-picked communities.
- Miniflux, a self-hosted RSS reader written in Go, installs in minutes with a single docker run command.
- RSS 2.0 has existed as a standard since 2002 and remains compatible with any modern reader.
- Going back to RSS doesn’t eliminate social networks, but it gives you back control over which sources you decide to follow.
What Happened
The alicegg.tech piece goes over four concrete cases of what it calls hijacked attention. On YouTube, the author argues, the platform knows you like cooking and art, but it still slips in clips about the stock market bubble, global warming, and the war in Iran in between: doomscrolling keeps the session open and multiplies ad clicks.
On Spotify, according to the essay, all it takes is opening a playlist and letting the algorithm build the session. The problem is what it inserts between songs: AI-generated tracks (“AI slop”) that the platform doesn’t have to pay any real artist for. On LinkedIn, a contact’s career update gets buried under opinions from strangers who, coincidentally, tend to work at whichever company Microsoft happens to hold a stake in at the time.
The fourth example points to Reddit: other users’ opinions about a product, which used to serve as a useful reference, now mix language model responses with Russian troll accounts, as the author describes it. The essay’s conclusion is that letting a third party decide what shows up on your screen is the same as handing over the key to your brain.
Context and History
The essay points out that the internet didn’t always work this way. Before recommendation algorithms became the norm, there wasn’t a single app deciding what to watch: there were a few dozen bookmarks to sites with a specific purpose, a video game news blog, a wiki for a 90s TV show, a tutorial site that updated rarely but well.
That doesn’t mean the old web was inherently a safe place. The author mentions Encyclopedia Dramatica and Rotten.com as examples of problematic content that still existed, but you had to actively go looking for it: nobody was suggesting it to you after a pancake recipe or a cat video.
Technically, the break coincides with the shutdown of Google Reader in 2013 and the subsequent mass adoption of the algorithmic feed on Facebook and Instagram during that decade: infinite scroll and engagement-based ranking replaced the chronological order of the sources you chose to follow. The intentional web, though, never fully disappeared. It’s still there, a bit buried under the corporate web, but you don’t have to dig too deep to find it.
Technical Details: How Attention Hijacking Works
A modern recommendation system doesn’t optimize for your long-term satisfaction, it optimizes for a proxy metric: watch time, click probability, or app return rate. The following pseudocode simplifies the logic most algorithmic feeds use:
def recompensa(contenido, usuario):
watch_time = predecir_tiempo_visto(contenido, usuario)
click_probability = predecir_click(contenido, usuario)
return watch_time * 0.7 + click_probability * 0.3
# The system picks the content with the highest expected reward,
# not the one that benefits the user most in the long run
siguiente_item = max(catalogo, key=lambda c: recompensa(c, usuario))That max() is the heart of the problem: any function that maximizes screen time will favor emotionally charged content (outrage, anxiety, morbid curiosity) over useful content, because that kind of content generates more app return on average. An RSS feed, by contrast, has no reward function: it returns entries in the order they were published, from the sources you added.
import feedparser
feeds = [
"https://alicegg.tech/feed.xml",
"https://news.ycombinator.com/rss",
]
for url in feeds:
parsed = feedparser.parse(url)
for entrada in parsed.entries[:5]:
print(f"{entrada.title} -> {entrada.link}")This script loops through two feeds and shows the last five entries from each, no ranking, no A/B testing, and no model deciding what to show first. The order is whatever the publisher defined, not whatever maximizes your screen time.
| Option | When to Use It | Advantage | Limitation |
|---|---|---|---|
| Algorithmic feed (YouTube, TikTok, X) | Discover new content effortlessly | Zero setup, there’s always something new | The selection criteria is engagement, not your actual interest |
| RSS / own reader (Miniflux) | Follow sources you’ve already chosen | Chronological order, no ads or tracking | Requires maintenance and manual curation |
| Email newsletter | Content curated by a single voice | Arrives without depending on a centralized feed | Clutters your inbox if you follow too many |
How to Get Started: Building Your Own Intentional Feed
The simplest technical step to take back control is to run your own RSS reader. Miniflux is a lightweight option, written in Go, built for self-hosting with Docker.
First you need Docker installed, depending on your operating system:
# Windows (PowerShell)
winget install Docker.DockerDesktop
# macOS
brew install --cask docker
# Linux (Debian/Ubuntu)
sudo apt install docker.io docker-compose -yWith Docker running, Miniflux spins up with a single container pointing to a Postgres database:
docker run -d --name miniflux -p 8080:8080 \n -e DATABASE_URL="postgres://miniflux:secreto@db/miniflux?sslmode=disable" \n -e RUN_MIGRATIONS=1 \n -e CREATE_ADMIN=1 \n -e ADMIN_USERNAME=admin \n -e ADMIN_PASSWORD=cambiala123 \n miniflux/miniflux:latestTo confirm the service is up and running, Miniflux itself exposes a health check endpoint:
curl -s http://localhost:8080/healthcheck
# returns: OKOnce you’re in, you can import dozens of sources at once with an OPML file exported from another reader, or add them manually using each site’s feed URL. For YouTube and Reddit, which don’t show RSS at first glance, extensions like Unhook (hides recommendations on YouTube) or going directly through old.reddit.com, which keeps the chronological order per subreddit, do the trick.
💡 Tip: almost any YouTube channel has an RSS feed at the URL https://www.youtube.com/feeds/videos.xml?channel_id=ID_DEL_CANAL, no API key needed.
Impact and Analysis
The essay’s argument connects to something measurable in product design: when a platform’s success metric is screen time, every design decision (autoplay, infinite scroll, push notifications) tends to push in the same direction. It’s not a one-off conspiracy by a single company, it’s the logical consequence of optimizing for that metric.
The difference with an RSS reader is structural: there’s no autoplay between different sources, no notification pushing you to come back, and the catalog of “what exists” is defined by whoever built the list of feeds, not a model trained to maximize retention. That comes at a cost: the intentional web isn’t infinite and doesn’t update with every click, and getting used to a slower internet takes time.
⚠️ Heads up: self-hosting your own reader means keeping the server updated and backing up the database; if that’s not feasible, a cloud reader (Feedly, Inoreader) solves the same problem without that operational burden.
flowchart TD
A["You"] --> B{"How do you access content?"}
B -->|"Algorithmic feed"| C["Platform decides the order"]
B -->|"RSS / bookmarks"| D["You decide the sources"]
C --> E["Optimized for engagement"]
D --> F["Chronological order, no ads"]What’s Next
The essay closes with a simple idea: the difference between the old web and today’s isn’t the available technology, it’s the user. Going back to reading blogs, following RSS feeds, and finishing that tutorial instead of continuing to scroll through shorts means adapting to a different pace, one where content isn’t infinite. Keeping up the habit, according to the author, is the only thing needed to change how you process information every day.
📖 Summary on Telegram: View summary
Try it yourself: install Miniflux with the Docker command from this article and add your favorite blog’s feed before you finish reading this.
Frequently Asked Questions
What Is the Tetris Effect Mentioned in the Essay?
It’s a psychological phenomenon where prolonged exposure to a stimulus (in the original experiment, the video game Tetris) causes the mind to keep processing and recognizing that pattern even outside the original context, for example while falling asleep or looking at everyday objects.
Do I Need to Know How to Code to Use RSS?
No. Any reader with a web interface, like Miniflux, Feedly, or NetNewsWire, lets you paste a site’s URL and start receiving its posts without writing any code. The Python script in this article is just an example of how it works under the hood.
Do RSS Feeds Completely Replace Social Media?
Not entirely. RSS solves the problem of consuming content from sources you already know, but it doesn’t replace social discovery or the direct interaction with other people that social networks do offer.
What’s the Difference Between Self-Hosting Miniflux and Using a Cloud Reader?
Self-hosting gives you full control over your data and no third party indexes what you read, in exchange for maintaining the server yourself. A cloud reader is simpler to get started with but depends on that company’s availability and policies.
Why Does Spotify Insert AI-Generated Content into Playlists, According to the Essay?
The author argues that this content doesn’t create any obligation to pay royalties to real artists, unlike a copyrighted song, which reduces the platform’s licensing costs.
References
- Attention is all you have: the original essay comparing the Tetris effect to algorithmic attention hijacking.
- Tetris effect (Wikipedia): description of the psychological phenomenon used as an analogy in the essay.
- Miniflux: official repository of the self-hosted RSS reader written in Go.
- RSS (Wikipedia): history and specification of the format, whose 2.0 version was published in 2002.
📱 Do you like this content? Follow @programacion on Telegram for daily tech content in Spanish: quick summaries, fresh content every day.
Imagen destacada: Foto de ThisisEngineering en Unsplash
0 Comments