On September 3, ChatGPT-6 was released. For the past two days, every tech news feed has been flooded with frenzied claims that “AGI is about to arrive.” Companies are running aggressive ads, marketing accounts are shouting “This is insane — humans are going to lose their jobs,” and even a simple version-number bump or capacity expansion is enough to draw crowds of worshippers.

Listen carefully: Stop the self-hype. The AGI you keep treating as divine is nothing more than the “emperor’s new clothes” jointly woven by tech capital and its users.

1. The Frenzy Over Mediocrity

Why do so many people think today’s AI is all-powerful? Because the vast majority of people’s daily work does not require real intelligence.

Writing formulaic official documents, fixing grammar, collecting existing materials, rewriting the same code for the hundredth time… These tasks are essentially just moving bricks and recombining within existing frameworks. AI, trained on massive data, can reassemble old knowledge into something smooth and fluent. That is good memory and high retrieval efficiency — it is not intelligence.

Those who find AI astonishingly powerful are merely excited by an extremely large old database. Their thinking has never gone beyond the boundaries of existing data, so when AI fluently spits out the clichés they already know, they mistake it for a miracle.

2. The Fatal Flaw: The Moment You Try to Explore a New Theory, AI Instantly Becomes a Complete Idiot

There is a simple way to tell whether something is true intelligence or just pretending to be smart: do not ask it to repeat old knowledge. Ask it to reason from a completely new concept.

Suppose you have finally discovered an unprecedented new theory — for example, that “1+1=3.” You pay for the latest ChatGPT-6 and ask it to abandon old ideas and help you further develop the academic implications of this new premise.

Guess how this “future deity” that has been hyped to the sky will respond? It immediately turns into an extremely stupid and rigid conservative, and delivers a classic “three-step idiocy”:

1. It lectures you with old theory
   You ask it to reason from “1+1=3.” Its first sentence is: “According to the Peano axioms, 1+1 strictly equals 2. Your assumption does not hold in traditional mathematics…”
   (I specifically asked you to reason from the new premise. Who are you lecturing with kindergarten knowledge?)
2. It changes the subject and forces everything back into the old framework
   When you forbid it from mentioning the old theory, it starts muddying the water: “In teamwork, the effect of 1+1 can be greater than 2, so 1+1=3. This is called synergy…”
   (Completely off-topic. It turns a rigorous logical axiom into corporate chicken soup.)
3. It floats into pure emptiness and piles up fancy words to pretend it is thinking
   When you pin it down and force it to write, it has no real logical content, so it starts spewing empty rhetoric: “1+1=3 reveals the emergent and nonlinear paradigms in the deeper dimensions of the universe, breaking the cognitive boundaries of binary opposition…”
   (Nothing but high-sounding empty vocabulary. Not a single solid logical inference. Pure floating nonsense.)

Whether it is GPT-5 or GPT-6, when faced with a genuine “from 0 to 1” conceptual rupture, it is simply a machine that cannot think.

3. Why Must It Behave Like an Idiot in Front of New Theories?

Because the underlying logic of current AI is probabilistic matching.

In its training data, “1+1=2” has appeared billions of times, while the “1+1=3” you just gave it is a brand-new disruptive concept. There is no real logical reasoning engine in its “brain.” There is only statistical inertia:

• The moment it starts generating, the underlying algorithm frantically pulls the words back onto the “old track” that has appeared billions of times.
• Your new theory is not treated as “innovation.” It is treated as “data noise” that needs to be corrected.

Current AI is essentially “the majority rules.” It has absorbed every human mental habit and conventional wisdom. By nature, it is anti-innovation.

4. No Matter How Big You Make a Plastic Bucket, It Still Cannot Hold Boiling Water

The “AGI breakthroughs” advertised by big companies are nothing more than making a plastic bucket thicker, larger, and sticking flashy labels on it.

It can swallow more tokens, gather more materials, and respond faster. These are only quantitative changes, never a qualitative leap. The core architecture has not changed at all. You can train a horse to run faster, but it will never grow a car engine. No matter how big you make the plastic bucket, the moment you pour boiling water (disruptive new logic) into it, it still deforms instantly.

Conclusion

Capital needs the AGI narrative to maintain valuations. Vendors need this new set of clothes to sell at high prices. People who only move bricks inside old frameworks need this new set of clothes to produce formulaic text quickly.

But please stop packaging an “efficient old-text collager” as “the superintelligence about to be born.”

While the entire internet is cheering for that non-existent glittering garment, the truth is very simple:

A rigid machine that cannot even explain a brand-new concept, that can only float into emptiness or drag everything back to old data the moment it meets a new theory, is still an entire civilization away from real AGI.