Where does the Vibe Coding bubble come from?

At a time when Vibe Coding is so hot that nearly everyone is talking about it, saying these things is bound to draw plenty of pushback.

But this article isn't meant to dismiss Vibe Coding — it is indeed a significant technological advancement. What I want to discuss is: why are so many people today developing nearly unlimited expectations for it?

Perhaps a few years from now, when this wave has passed, people digging up today's discussions will see things differently.

---

Where Does the Vibe Coding Bubble Come From?

Fake it until you make it: Silicon Valley culture

Silicon Valley has long had a famous saying:

Fake it until you make it.

It was originally a culture that encouraged entrepreneurs to boldly experiment and quickly validate ideas — and it's precisely this culture that has nurtured countless innovative companies.

But on the other hand, this culture can easily devolve into another extreme — continuously amplifying technology's capabilities, even cashing in on the future ahead of time.

So we often see narratives like this:

• AI will replace all programmers;
• SaaS is about to die;
• A single prompt can complete the entire software development process.

Marketing has no floor, hype has no ceiling. There's nothing they can't claim to do, and no trick they won't pull.

Two characters worth mentioning in particular:

• Sam Altman of OpenAI: from "AGI is a few years away" to "a few months away," to constantly emphasizing superintelligence (ASI), trillion-dollar compute investments, and the Stargate project — the narrative keeps escalating. Every time the market starts adapting to the previous story, an even bigger vision appears.
• Dario Amodei of A-something: he puts out new AI capability predictions almost every other week — SaaS is dead, Myous is too dangerous, we shouldn't let people use it, everyone should pause AI research for safety — and then a few days later, he releases Fable 5.

Even Old Huang has mocked this, wondering how Dario ever became CEO in the first place.

These views aren't entirely without basis, but they often ignore the vast number of real-world engineering problems, packaging capabilities that belong years or even decades in the future as if they were already facts today.

---

The Gell-Mann Amnesia Effect

There's a famous psychological phenomenon called the Gell-Mann Amnesia Effect.

It refers to this:

When you read news about a field you know well, you'll find it riddled with errors; but when you flip to a field you don't understand, you somehow trust the same media outlet again.

In a sense, LLMs are becoming the information gateway of the new era.

The problem is that the fields a person truly masters are actually very limited.

For familiar domains, it's easy to spot errors in a model's responses; but when faced with the other ninety-plus unfamiliar domains, we often lack the ability to judge and can only default to assuming it's correct.

As a result, people gradually develop an illusion:

If it can answer things I don't understand, then doesn't it know everything?

Compared to traditional media, LLMs have nearly unlimited capacity for information production, which further amplifies this psychological effect.

In the end, people overestimate the model's capabilities while underestimating the complexity of the problems behind them.

---

Capital manipulating the narrative

Today, our access to information increasingly relies on short-video platforms, community platforms, and various AI-generated content.

Although information sources are more fragmented than in the past, the platforms that truly control content distribution remain highly concentrated.

Platforms naturally amplify the most shareable topics, and capital prefers telling the fastest-growing, most attention-grabbing stories.

So a technology that should be rationally discussed easily turns into a new narrative that everyone chases.

Vibe Coding is not the first concept to go through this process, and it won't be the last.

Vibe Coding: The Most Successful Con Job on Clueless Bosses

Many bosses genuinely lack competence. A lot of people earned their first pot of gold not through technical, product, or management skills, but by riding the dividends of their era, policy tailwinds, information asymmetry, channels, resources, connections, or even speculative arbitrage.

They did make money and gained wealth and status, but their fundamental technical understanding isn't much higher than the old man at the village entrance — they just had better luck, more opportunities, and stood at the right place at the right time.

What they see is:

• AI writing a webpage in minutes;
• Completing a demo in an hour;
• Social media constantly showing cases of "one person building an entire product."

So a natural conclusion follows:

If a demo can be done in minutes, why does the real product still need so many developers?

But a demo and a truly maintainable software product are two completely different things.

Many people outside the industry never get the chance to encounter the genuinely hard problems, such as:

• Long-term maintenance;
• System architecture;
• Data consistency;
• Security;
• Performance optimization;
• Multi-person collaboration;
• Continuous evolution after launch.

So what they see is AI dramatically lowering the barrier to starting development, yet they mistakenly believe AI has also reduced the complexity of software engineering itself.

---

No Silver Bullet

Decades ago, the software engineering field already proposed a classic view:

There is no silver bullet in software engineering.

Today, that statement still holds.

Vibe Coding has massively accelerated demo generation and given individual developers productivity they couldn't have imagined before.

But from a demo to a truly deliverable, maintainable, and scalable product, there remains a huge amount of engineering work in between.

AI has shortened coding time, but it hasn't eliminated these engineering problems.

So I prefer to understand Vibe Coding as a new programming abstraction.

It's not the end of software engineering — it's more like a programming language one level higher than Python — a new way of programming that uses natural language as its interface.

Complexity hasn't disappeared; it's just been hidden

Historically, this has happened many times before.

When Python appeared, it dramatically lowered the barrier to programming.

More and more people started using Python to build websites, automation tools, data analysis programs, and even desktop applications.

But more than twenty years later, Java and C++ still firmly hold their ecosystems, and new languages like Go and Rust have since emerged.

Why?

Because complexity has never truly disappeared.

Python hides a large amount of low-level detail from developers — memory management, compilation pipelines, boilerplate code, and so on.

These complexities weren't solved; they were shifted into the runtime, the interpreter, the frameworks, and various underlying implementations.

Vibe Coding works the same way.

It further hides complexity behind the models and agents.

Work that previously required programmers to do themselves is now handed to AI for automatic generation.

What used to require understanding APIs, frameworks, and syntax is now just a sentence of natural language.

But the problems a real software system faces — changing requirements, architecture design, data consistency, security, performance, maintenance costs, team collaboration — still exist.

Complexity doesn't vanish into thin air.

It only keeps migrating to deeper layers.

And the evolution of software engineering has essentially always been a process of constantly finding new abstractions to hide complexity downward.

The Future Is Already Here

Vibe Coding is just the next stop on this evolutionary path, not the destination.

We can't go back to the farming era of manual typing, but the Vibe Coding of the new era is definitely not what today's hotshot AI self-media influencers and bandwagon-jumping bosses are describing.