How absurd! OpenAI's pixel-level distillation of a Chinese company has been completely debunked by frame-by-frame comparison.

Remember that viral long article that blew up all over the internet a few days ago?

On September 7, Chief Scientist Jakub Pachocki published a 10,000-word long article titled *An Alien Mind*, telling the whole world: We are building a kind of intelligence that humans cannot fully understand.

The moment this article was released, it immediately caused a huge stir across the entire internet.

However, after the incident broke out, a company based in Suzhou, China fell into silence...

The more they read, the more they felt something was wrong.

Wasn't that grand and profound conceptual framework described by Pachocki, the flying particle effects on the GPT-6 Astra official website page, and the core expressions about "self-evolution", "cognitive-like neuroscience" and "value alignment" all content from their own official website?

A few days ago, a mysterious robot video went viral all over the internet, drawing huge attention in the industry.

This model that has amazed countless people, codenamed MVP, does not come from Silicon Valley, but from a genuine Chinese company: Enjoy Tech.

Originally, this company planned to continue polishing the model and keep the suspense going.

But in the past 48 hours, a series of "coincidences" that happened on the other side of the ocean made them decide: No more hiding, we're laying our cards on the table, and officially claimed the ownership on September 10!

If they didn't speak up soon, the technical concepts and model ideas that they had already verified months ago might be "legitimately" taken away by OpenAI!

A Striking "Pixel-Level" Duplication

First of all, let's take a look at the timeline.

More than a month ago (July 2026), Enjoy Tech officially launched the official website of the Aether large model (aether.cn), and simultaneously released product concept explanations and DEMO videos, fully disclosing everything from model architecture to technical routes to core concepts.

A few weeks ago, the founder of Enjoy Tech met face-to-face with many industry peers in Silicon Valley, USA, and publicly demonstrated their core conceptual framework about "alien visitors": this "non-human agent / alien mind" is a brand new form of intelligence that grows without following human cognitive logic.

September 3 (US time), OpenAI suddenly released the official website of GPT-6 Astra.

September 6, Pachocki published the 10,000-word long article *An Alien Mind*.

Industry insiders found that the content released by OpenAI this time, from UI visuals and core concepts to technical philosophy and alignment strategies, is "highly similar" to the content released by Enjoy Tech more than a month ago.

Next, let's compare frame by frame to see exactly who is "drawing inspiration" from whom.

First question: Why is the official website design so similar?

Open aether.cn, the official website of Enjoy Tech, you will see dynamic particle effects floating on a dark background, flowing like a star river, full of a strong sense of technology, conveying the signal of "intelligence on a cosmic scale".

This page was launched as early as July. All design elements, interaction logic and visual language can be publicly accessed and verified.

And "particle effect" is the core concept of Enjoy Tech's design.

Then open the official website of OpenAI's GPT-6 Astra — dark background, flowing particle effects, starlight converging into the shape of "6". The visual language, atmosphere building and design intuition are almost identical.

By comparison, it looks like OpenAI directly copied someone else's web page.

This page was launched on September 3.

Some people may say that dark background plus particle effects are not uncommon in technology websites.

That makes sense, but the biggest doubt this time lies in the details: the motion logic of particles, the overall visual narrative rhythm of the page, and the hierarchical structure of information presentation, this pixel-level "reproduction" is hard to explain as a coincidence.

Second question: Why are the core concepts highly overlapping?

If the website design can barely be explained as aesthetic convergence, the following things are even more weird.

A few weeks ago, the person in charge of Enjoy Tech clearly put forward the persona direction of "alien mind" during exchanges in Silicon Valley (even humorously proposed in internal discussions to "set it as an alien that landed in China combined with Eastern philosophy").

This statement is their definition of the essence of the Aether large model: this kind of intelligence is not a product of imitating humans, but a brand new, alien form of mind, which is driven by energy, based on physical laws, grows and evolves autonomously.

A few weeks later, Jakub Pachocki, Chief Scientist of OpenAI, directly used *An Alien Mind* as the title to publish the 10,000-word long article.

Not only the titles collide, but the core connotations also overlap.

Pachocki wrote in the article: "AI is more likely to 'grow' rather than be 'designed'. It is the product of repeating simple optimization steps on massive computing power. The final system is extremely complex, and we can discover various emerging micro-mechanisms inside it through research similar to neuroscience, but we can never fully understand its overall operation mode."

However, on Enjoy Tech's official website, there was such a statement as early as a month ago: their model architecture is bionic, "building a silicon-based traveler according to the structure of the human brain — cerebrum, cerebellum, cortex, spinal cord", using neuroscience methods to simulate every component of the human brain structure and consciousness.

The core connotations are so similar that it seems very likely to be a case of "cultural export".

Third question: Why do the technical concepts correspond layer by layer?

Next, the more you dig, the more shocking you will find.

For example, the similarity of the evolution path of machine intelligence.

Enjoy Tech believes that energy is more essential than language, so it is the first in the world to introduce energy into the training of embodied models. After one and a half months of mathematical derivation, when they finally saw the result expressed as a single line of formula, their shock was no less than the surprise when people first saw the theory of relativity express the universe with a single line of formula.

The technical core of the Aether large model is "energy-driven" — instead of directly predicting actions, it unifies task completion, physical constraints, and dynamic consistency into a continuous energy landscape, and completes reasoning and decision-making under the energy gradient.

However, in Pachocki's article, there is also a discussion about the unified optimization goal and internal driving mechanism.

For example, the concept of RSI.

One of the core highlights of Enjoy Tech's Aether large model is "self-evolution" — through the energy-driven Evolution Manager, it achieves multi-level self-evolution from the parameter level to the model level.

They have enabled robots to achieve "independent control over their own development without any human intervention".

However, Pachocki also discussed "Recursive Self-Improvement (RSI)" at length in the article, that is, the AI model can independently optimize itself and dominate its own development process, and also mentioned that AI will in turn optimize the computing power infrastructure itself.

This is such a coincidence, it can be said that OpenAI has helped Enjoy Tech pre-announce the optimization of computing power.

There is also the emphasis on generalization ability.

The official website of the Aether large model also takes generalization ability as a key introduction — "Our generalization is not a VLA in disguise that searches according to a map, but is driven by instinctive awe of physical laws." Even the underlying principles are fully explained.

Pachocki's original text is: We have spent a lot of time trying to figure out how capabilities generalize, so related work on RSI and automated alignment research is more urgent.

The most coincidental thing is that in the deep-water "alignment" issue, the theoretical frameworks of the two sides are completely identical!

In the DEMO, Enjoy Tech creatively splits alignment into "goal alignment" (strive to complete instructions) and "value alignment" (be reasonable in fuzzy adversarial environments, uphold honesty and kindness).

OpenAI's article also emphasizes that it is "necessary to distinguish between goal alignment and value alignment", even the underlying logic is pixel-level similar.

In addition, there is the "inexplicability" of cognitive-like neuroscience.

Enjoy Tech's model builds the model architecture according to the structure of the human brain (neuroscience), and uses neuroscience methods to simulate every part of the human brain structure (and consciousness). This is the essence of the model — to create a silicon-based life with a complete cerebrum-cerebellum-cortex-spinal cord structure.

For this reason, their web page design is designed according to the neuroscience of the human brain.

OpenAI's long article states that "AI 'grows' — we can gain insights into various emerging micro-mechanisms inside this system through research similar to neuroscience, and the overall operation mode of this system can never be fully understood and described by us."

It can be said that this is a pixel-level reproduction of the core essence of Enjoy Tech's model.

After reading these frame-by-frame comparisons, what do you think?

It is certain that this is not a collision of one or two concepts, but a high degree of overlap of the entire technical narrative framework!

OpenAI's Alien Mind, the Chinese Team Has Already "Built" It

Faced with OpenAI's overwhelming theoretical promotion, Enjoy Tech only took one action to respond