Is this the beginning of the end for YouTube? (yes I'm serious)
wono@wono_strategy
10 views
Oct 22, 2025
~4 min read
Advertisement
1
Is this the beginning of the end for YouTube? (yes I'm serious)
2
Where is content creation heading with AI?
Level 1: You won't be replaced by AI, but by someone using AI
Level 2: Competitive advantages are collapsing
Most people stop at level 2, so here's level 3:
Level 1: You won't be replaced by AI, but by someone using AI
Level 2: Competitive advantages are collapsing
Most people stop at level 2, so here's level 3:
3
AI is about to break recommendation systems.
Historically, all attention breakthroughs (printing press, radio, tv, internet...) expanded the total "territory of attention" (though more distribution).
More content (supply), but also new audiences reached (demand).
Historically, all attention breakthroughs (printing press, radio, tv, internet...) expanded the total "territory of attention" (though more distribution).
More content (supply), but also new audiences reached (demand).
4
But for the first time in history, this territory is about to be seriously challenged:
The constraints (time, effort, cost) that stabilized the attention economy are collapsing.
Algorithms were built in an environment where producing content had these natural constraints.
The constraints (time, effort, cost) that stabilized the attention economy are collapsing.
Algorithms were built in an environment where producing content had these natural constraints.
5
Attention doesn't scale, content production now does (at an unprecedented scale).
Humans can't bypass time, while AI just requires more GPUs.
So far, recommendation algorithms have been the gatekeepers to filter that content.
But we now have 2 (big) problems.
Humans can't bypass time, while AI just requires more GPUs.
So far, recommendation algorithms have been the gatekeepers to filter that content.
But we now have 2 (big) problems.
6
1) Algorithms are not perfect (far from it), they make mistakes.
2) They require constant human feedback (attention)
Human attention is finite, AI content is virtually infinite.
2) They require constant human feedback (attention)
Human attention is finite, AI content is virtually infinite.
7
Algorithms could handle human content because of those natural constraints, BUT:
- Each new video has a non-zero potential for error
- Corrections depend on a finite amount of reliable (human) feedback
If the content supply grows too fast, the absolute number of errors increases fast as well (EVEN if the algorithm error rate decreases.)
- Each new video has a non-zero potential for error
- Corrections depend on a finite amount of reliable (human) feedback
If the content supply grows too fast, the absolute number of errors increases fast as well (EVEN if the algorithm error rate decreases.)
8
In other words, it doesn't matter how good the algorithm gets:
- its errors scale with more content while its accuracy doesn't
Here's a simple simulation showing that there's an inevitable breaking point in terms of what the system can handle (even if it gets better).
- its errors scale with more content while its accuracy doesn't
Here's a simple simulation showing that there's an inevitable breaking point in terms of what the system can handle (even if it gets better).
0:37
9
Understand that I'm not saying AI is the problem here.
The problem is that the current recommendation systems can't filter what it was never meant to face.
A dam built for a river can't suddenly hold back an entire ocean.
The problem is that the current recommendation systems can't filter what it was never meant to face.
A dam built for a river can't suddenly hold back an entire ocean.
10
When content supply is manageable, you get some wrong recommendations here & there, buried under better options.
But when the same volume explodes, so does the recommendation error.
Eventually, there's more mud in the water than water and the system can't tell the difference.
But when the same volume explodes, so does the recommendation error.
Eventually, there's more mud in the water than water and the system can't tell the difference.
11
This is because when AI content grows faster than human feedback, every recommendation adds layers of uncertainty:
1st order error: a wrong recommendation
2nd order: misjudging that error
etc
Until eventually: errors in the correction process itself feed into each other.
1st order error: a wrong recommendation
2nd order: misjudging that error
etc
Until eventually: errors in the correction process itself feed into each other.
0:24
12
Another problem is how the current incentive structure pushes the system toward its own breaking point:
When one viral video made effortlessly can pay for hundreds of flops, spam becomes the rational move.
The more you upload, the better your odds of earning money.
When one viral video made effortlessly can pay for hundreds of flops, spam becomes the rational move.
The more you upload, the better your odds of earning money.
13
Under this transition of uncertainty, these are the 3 most important market dynamics principles to keep in mind:
1) Nature selects through elimination
2) Adaptation happens through chaos
3) Stability only exists under constraints
1) Nature selects through elimination
2) Adaptation happens through chaos
3) Stability only exists under constraints
14
We are entering an era where:
- Algorithms are structurally built on a constraint that can now be bypassed (time, effort, cost)
- AI can clone & flood faster than platforms can regulate
- Content grows faster and faster while attention stays the same
- Algorithms are structurally built on a constraint that can now be bypassed (time, effort, cost)
- AI can clone & flood faster than platforms can regulate
- Content grows faster and faster while attention stays the same
15
Remember that a system only reveals its limits once it breaks.
And when it breaks, everything built on the assumption of its stability gets stress-tested next.
And when it breaks, everything built on the assumption of its stability gets stress-tested next.
16
Whatever your business model is:
- ad revenue
- sponsors
- courses
- etc
All of that was built assuming human effort limited your competition.
Expect AI to stress test whether that model still works when anyone can copy you or flood your attention market at near 0 cost.
- ad revenue
- sponsors
- courses
- etc
All of that was built assuming human effort limited your competition.
Expect AI to stress test whether that model still works when anyone can copy you or flood your attention market at near 0 cost.
17
One thing is for sure:
Markets adapt the only way they know: through elimination, by destroying what no longer works.
Most people look at how AI can affect content creators, but how many realize it could kill the platforms themselves?
Markets adapt the only way they know: through elimination, by destroying what no longer works.
Most people look at how AI can affect content creators, but how many realize it could kill the platforms themselves?
18
- New (or old) actors will discover new competitive advantages that we can't imagine yet (as we're still in the old paradigm)
- Those who won't adapt (the majority) will die
AI could very well be the final boss of the recommendation systems as we know them today.
- Those who won't adapt (the majority) will die
AI could very well be the final boss of the recommendation systems as we know them today.
19
What will happen in the future?
I don't know.
By the time we know what works, it will be too late to prepare for it. The new rules will be written by those who survived long enough to see them emerge.
Thank you for your attention.
I don't know.
By the time we know what works, it will be too late to prepare for it. The new rules will be written by those who survived long enough to see them emerge.
Thank you for your attention.