It happened to me again recently.
I was reading an interesting, slightly longer piece on LinkedIn. No selfie in front of a conference banner, no motivational quote about “leadership,” but something that actually needed a few minutes of attention.
Then something terrible happened: real life got in the way.
I put my phone down, did something else, and came back a few minutes later to keep reading.
Bam.
Feed refreshed.
Article gone.
And good luck finding it again.
Instagram and Facebook are even worse. Instead of finding that one interesting post, you’re immediately served another selfie. Someone who, after forty-five minutes of thinking, has discovered that “people are the most important asset.” A video of Emirates First Class. A watch. A Hermès or Louis Vuitton bag that would cost anyone else roughly a year to a year and a half of a typical salary.
But the piece you actually wanted to read? Too bad. Good luck with that.
Maybe this isn’t a bug
I don’t know whether that refresh was deliberately designed to produce this exact effect. That would be too easy an accusation.
But the result fits remarkably well within the larger system.
Social media feeds aren’t digital newspaper pages. They’re recommendation engines that are constantly recalculating.
LinkedIn says this fairly openly. Its algorithm considers hundreds of signals: what you view, respond to, or skip, who you interact with, and even how much time you spend on content.
AI then analyzes large volumes of previous interactions to predict what you’ll probably find relevant or interesting.
That sounds helpful.
Until you ask:
relevant to what?
Because that’s the fundamental problem.
An algorithm can optimize for almost any measurable outcome.
How much someone learns. Whether someone actually finishes an article. Whether someone understands something better. Concentration. Professional development.
Or simply:
how much longer will someone stay on the platform?
That choice isn’t technical.
It’s a business decision.
From engagement to intent
For me, that’s also the interesting question around AI.
We’re getting better and better at predicting where someone’s attention will go.
But why don’t we use that technology to understand where that person actually wants to go?
Engagement asks:
How do we keep Thierry here?
Intent asks:
What did Thierry come here for?
That seems like a small difference.
It’s enormous.
The first question naturally leads to infinite scrolling, autoplay, notifications, engagement loops, and increasingly sophisticated recommendations.
The second could produce a completely different kind of technology.
Suppose I open LinkedIn at 8:30 a.m.
The system knows I’m interested in IT leadership, technology, organizational change, and interim assignments.
AI could then say:
“Here are the seven posts from your network that are likely to be truly relevant to you today. Two job opportunities, one good article on AI governance, an interesting leadership discussion, and three updates from people you regularly stay in touch with.”
And then:
You’re caught up.
Done.
No bottomless feed.
No endless stream of random stimulation.
No algorithmic rabbit hole because I looked at something about a Lamborghini once and apparently won’t need anything else for the next three weeks.
Maybe a truly human-centered AI would even say:
“That article you stopped halfway through yesterday? You had about three minutes left to read. Here it is.”
Revolutionary.
A computer that remembers what I wanted.
Instead of only what the platform wanted from me.
Welcome to the attention industry
The problem, of course, is that almost every major social platform ultimately competes for one scarce resource:
- YOUR ATTENTION
Not your time in some abstract sense.
Your eyes. Your curiosity. Your boredom. Your irritation. Your need for social validation.
Every extra minute is measurable.
Every scroll is measurable.
Every click is measurable.
Every second you linger on a post can become a signal.
And wherever attention can be monetized, an economic incentive to produce more attention follows naturally.
In 2025, Meta earned more than $196 billion from advertising out of roughly $201 billion in total revenue.
That isn’t a detail of the business model.
That is the business model.
We’ve created a peculiar situation.
Some of the world’s best software engineers, data scientists, behavioral scientists, and AI specialists work on systems that can predict with astonishing precision what will hold our attention for a few more seconds.
And then we use all that intellectual and technological firepower for…
another Reel.
Another suggested post.
Another ad.
Just one more.
The unchecked growth of advertising
And then there are ads.
I deliberately use the term unchecked growth.
In medicine, we call uncontrolled cell growth cancer: cells that keep multiplying and eventually impair the functioning of the organism.
Obviously, this is not equating advertising with the human suffering caused by cancer. Anyone who has experienced it knows that reality is on a completely different level. Unfortunately, I’ve seen it up close several times myself.
But as a metaphor for what advertising has done to parts of the internet, the comparison remains uncomfortably apt to me.
- One ad.
- Two ads.
- Sponsored posts.
- Promoted content.
- Suggested for you.
- A commercial influencer who doesn’t technically appear to be an ad.
- A video starts.
- An ad before it.
- Another halfway through.
- A cookie banner.
- A newsletter popup.
- “Download our app!”
- A push notification.
Somewhere among all those commercial metastases is still the information you originally came for—if you can find it underneath the popups and overlays.
I now run AdGuard.
Not because I’m against advertising on principle. A newspaper, website, or service needs a way to pay its bills.
But there’s a difference between advertising funding the content and using content as an excuse to show advertising.
Across huge parts of the internet, we’ve lost that distinction.
AI can do the exact opposite
And that’s what actually interests me.
AI doesn’t have to be an attention-stealing machine.
It could become a filter against digital junk.
A system that reduces twenty thousand possible distractions to the ten that matter today.
That remembers what you were doing.
That helps you focus.
That doesn’t just predict what you’ll click, but understands what you’re trying to achieve.
That distinction is essential.
Because once you optimize for intent rather than engagement, you get an entirely different product.
“This is what you wanted to know.”
“This is what you missed.”
“These three things matter.”
“The rest is noise.”
“See you tomorrow.”
From an advertising-revenue perspective, that probably sounds like an exceptionally bad product.
From a human perspective, perhaps an exceptionally good one.
Perhaps that’s exactly the problem
We like to complain about our phone addiction.
As if billions of people happened to develop insufficient discipline at the same time.
I find that too easy.
We’re using products deliberately made extremely good at predicting, selecting, and delivering stimuli.
Up against that is a human brain that evolution didn’t exactly design to resist a machine learning model with billions of data points every twenty seconds.
The European Digital Services Act is cautiously beginning to acknowledge this problem. Among other things, very large platforms must offer users an alternative to recommendation systems based on profiling.
But perhaps we need to think one step further.
Not just:
“Give me a less manipulative algorithm.”
But:
“Give me an algorithm that demonstrably works for me.”
That protects my attention instead of exploiting it.
That treats my goal as more important than its revenue per user.
And perhaps even occasionally dares to say:
There’s nothing more for you here. Go do something fun in real life.
Although, of course, real life shouldn’t get in the way too often.
Before you know it, your feed will refresh again.
And now I’m actually curious:
Did you really make it to the end of this article? Let me know in the comments.
Originally published on LinkedIn. View all blog posts.