I think we're currently going through another version of the same illusion I've seen before.
A few years ago, I was spending a lot of time learning hacking, especially low-level stuff.
I already knew how to program. I could build normal websites and applications. I understood software reasonably well.
But when I started going deeper into exploitation, I realized something:
Knowing how to write software isn't the same thing as understanding how software actually works underneath.
If I wanted to find serious vulnerabilities, I couldn't just keep learning higher-level programming.
I needed to understand machine language, assembly, memory, CPU architecture, calling conventions, operating system internals, etc.
At some point I realized that if I didn't understand the lower layers, there was a ceiling to what I could actually discover.
And this is exactly why I think vibe coding is being misunderstood.
And I wish For everyone can code their app without knowing coding to publish their apps and make money off it. but from what I've seen yes sometimes, but still some way for ai to improve.
That's a massive difference.
Today, you can tell an AI:
That's insane.
Ten years ago, you would have needed significantly more technical knowledge and time just to get something close to that.
But then you start using the application.
And this is where things get interesting.
You discover that:
And I think this is the fundamental problem with the "vibe coding will make everyone a millionaire" narrative.
They're valuable because they understand systems.
They understand why something should be built one way instead of another.
They understand tradeoffs.
They know what can break.
They know which assumptions are dangerous.
They know what information one component needs from another.
They understand the architecture underneath the interface.
And most importantly:
They can recognize when something is wrong even when the application appears to work.
This is very similar to what I experienced with low-level hacking.
I could write software without knowing assembly.
But I couldn't expect to become really good at finding certain classes of vulnerabilities without understanding what was happening underneath the abstraction.
Vibe coding creates a similar abstraction.
It lets you operate at the surface.
And the surface is becoming incredibly powerful.
But the deeper layers haven't disappeared.
AI is extremely powerful.
A software engineer who knows what they're doing can now produce an absurd amount of software.
A person who previously needed a small team to build an MVP can sometimes do a huge amount of it alone.
AI basically turned a strong developer into something that looks like a small software company.
And that's probably going to continue.
But there's an important distinction:
AI amplifies existing ability much more effectively than it creates engineering judgment from nothing.
Current AI systems are fundamentally pattern-based systems.
They can generate extremely impressive solutions because they've learned from enormous amounts of existing information.
But generating something that looks correct is not the same as having a deep understanding of why that system should exist, what its constraints are, what the business actually needs, and which tradeoffs matter.
That's where human experience still matters enormously.
But probably not in the way people imagine.
The winners won't necessarily be the people who can generate the most code.
They'll be the people who can combine AI with something AI doesn't automatically give them:
experience.
A person who understands software deeply + AI can move ridiculously fast.
A person who understands a specific industry + AI can build things that previously required an entire development team.
A person with distribution + AI can launch and iterate at an insane speed.
But someone with no technical knowledge, no domain knowledge, no distribution and no understanding of customers?
Giving that person an AI coding agent doesn't magically give them all four.
It just makes them faster at producing software.
And producing software was never the same thing as building a business.
Maybe the real revolution isn't:
A few years ago, I was spending a lot of time learning hacking, especially low-level stuff.
I already knew how to program. I could build normal websites and applications. I understood software reasonably well.
But when I started going deeper into exploitation, I realized something:
Knowing how to write software isn't the same thing as understanding how software actually works underneath.
If I wanted to find serious vulnerabilities, I couldn't just keep learning higher-level programming.
I needed to understand machine language, assembly, memory, CPU architecture, calling conventions, operating system internals, etc.
At some point I realized that if I didn't understand the lower layers, there was a ceiling to what I could actually discover.
And this is exactly why I think vibe coding is being misunderstood.
And I wish For everyone can code their app without knowing coding to publish their apps and make money off it. but from what I've seen yes sometimes, but still some way for ai to improve.
Vibe coding removed the barrier to writing code.
It did NOT remove the barrier to building a good product.That's a massive difference.
Today, you can tell an AI:
And 10 minutes later, you might actually have something running.
That's insane.
Ten years ago, you would have needed significantly more technical knowledge and time just to get something close to that.
But then you start using the application.
And this is where things get interesting.
You discover that:
- some features don't actually work properly
- edge cases weren't considered
- permissions are wrong
- data isn't handled correctly
- error states are missing
- the UI looks complete but the underlying system isn't
- some features are essentially decorative
- different parts of the application don't really understand each other
- changing one thing breaks another thing
And I think this is the fundamental problem with the "vibe coding will make everyone a millionaire" narrative.
Software is not just code.
A good software engineer isn't valuable because they can type Python or JavaScript quickly.They're valuable because they understand systems.
They understand why something should be built one way instead of another.
They understand tradeoffs.
They know what can break.
They know which assumptions are dangerous.
They know what information one component needs from another.
They understand the architecture underneath the interface.
And most importantly:
They can recognize when something is wrong even when the application appears to work.
This is very similar to what I experienced with low-level hacking.
I could write software without knowing assembly.
But I couldn't expect to become really good at finding certain classes of vulnerabilities without understanding what was happening underneath the abstraction.
Vibe coding creates a similar abstraction.
It lets you operate at the surface.
And the surface is becoming incredibly powerful.
But the deeper layers haven't disappeared.
AI changed the economics of software development.
This is the part where I actually agree with the hype.AI is extremely powerful.
A software engineer who knows what they're doing can now produce an absurd amount of software.
A person who previously needed a small team to build an MVP can sometimes do a huge amount of it alone.
AI basically turned a strong developer into something that looks like a small software company.
And that's probably going to continue.
But there's an important distinction:
AI amplifies existing ability much more effectively than it creates engineering judgment from nothing.
Current AI systems are fundamentally pattern-based systems.
They can generate extremely impressive solutions because they've learned from enormous amounts of existing information.
But generating something that looks correct is not the same as having a deep understanding of why that system should exist, what its constraints are, what the business actually needs, and which tradeoffs matter.
That's where human experience still matters enormously.
So will vibe coding create millionaires?
Absolutely.But probably not in the way people imagine.
The winners won't necessarily be the people who can generate the most code.
They'll be the people who can combine AI with something AI doesn't automatically give them:
experience.
A person who understands software deeply + AI can move ridiculously fast.
A person who understands a specific industry + AI can build things that previously required an entire development team.
A person with distribution + AI can launch and iterate at an insane speed.
But someone with no technical knowledge, no domain knowledge, no distribution and no understanding of customers?
Giving that person an AI coding agent doesn't magically give them all four.
It just makes them faster at producing software.
And producing software was never the same thing as building a business.
Maybe the real revolution isn't:
It's:
And that's a much more interesting revolution.
Last edited: