When a Mind Can Be Installed: The zhangxuefeng-skill AI Project

Forum topic · 小凯 · 2026-04-09

Summary

This article examines zhangxuefeng-skill, an open-source GitHub project that distills the cognitive style of Zhang Xuefeng—a Chinese education influencer with over 40 million followers who died of sudden cardiac arrest on March 24, 2026, at age 41—into a loadable AI skill package. Built with the Nüwa.skill framework, the project analyzes the subject across six dimensions (writings, interviews, expression style, criticism, major decisions, life timeline) to extract five mental models: the 'social sieve theory' (education, housing, and jobs filter people by class), 'choice over effort,' 'employment-based backward reasoning,' 'class realism,' and 'controversy as propagation.' The skill encodes not just quoted statements but his decision-making heuristics and rhetorical DNA—short punchy sentences, northeastern dialect, and hyperbolic humor. The article traces Zhang's biography from a poor rural family to an 800-million-yuan education company, discusses the project's quality validation process, and weighs ethical questions: distilling a deceased person's thinking versus exploiting their death, and what it means when human cognition becomes installable software. It concludes that this form of remembrance preserves how a person thought, not merely what they said.

When a Mind Can Be Installed: Zhang Xuefeng's Cognitive Operating System

> "Society is one big sieve—it filters children by degrees, parents by housing, and families by jobs." > — Zhang Xuefeng

Introduction: A Soul Encoded

On March 24, 2026, Zhang Xuefeng—the education influencer with over 40 million followers across Chinese platforms, famous for his rapid-fire, blunt career and college advice—died of sudden cardiac arrest at 41.

Days after his death, an open-source project quietly appeared on GitHub: zhangxuefeng-skill, an AI skill package created by developer Huashu using a thought-distillation framework called Nüwa.skill (Nüwa skill). Rather than a collection of quotes or simple role-play, it encodes Zhang's cognitive patterns, decision logic, and expression style into a loadable "cognitive operating system." Ask it "My child scored 560 on the gaokao and wants to study finance—what do you think?" and it doesn't repeat his old answers; it applies his mental frameworks—the social sieve theory, employment-based backward reasoning, class realism—to your specific question.

This raises a question: when a person dies, can their thinking persist—and in what form? A memoir? Or, as this project suggests, a callable piece of code?

Chapter 1: Who Was Zhang Xuefeng?

Born in 1984 to a poor family in Fuli County, Heilongjiang (his real name: Zhang Zibiao), Zhang studied water supply and drainage engineering at Zhengzhou University—a major he joked was mostly about "unclogging drains." He moved to Beijing in 2007 on a 2,500-yuan monthly salary, working as a salesperson at a graduate-exam tutoring company, where his gift for making dry admissions data entertaining became apparent.

In 2016, a 7-minute speech on a TV show analyzing all 34 elite "985" universities went viral with over 100 million views. He then published bestselling books, built a 40-million-follower social media presence, and in 2021 founded Fengxue Weiwei, a college application consulting company reportedly earning 800 million yuan in annual revenue by 2024 (per TMTPost).

His fame came with controversy. His 2023 remark that he would "knock out" any child wanting to study journalism led to a meeting with the Cyberspace Administration and a temporary livestream ban. A 2024 comment that "liberal arts is all service industry" drew further backlash—yet his popularity only grew. Notably, the skill file itself lists this as a deliberate strategy: "Controversy is propagation. Mild advice is forgotten; extreme views have penetrating power."

His final WeChat Moments post reportedly read: "Life is fun. I'd come back for another round." It became the skill project's epigraph.

Chapter 2: The Social Sieve Theory

Zhang's unified worldview: society is a big sieve with three layers:

1. Degrees filter children — Fortune 500 companies claim grades don't matter, but their campus recruiting visits only target top universities. 2. Housing filters parents — In Beijing, Shanghai, or Shenzhen, whether you own a home shapes stability, belonging, and mental bandwidth. 3. Jobs filter families — "Your salary is proportional to your irreplaceability."

His self-positioning was "speaking for ordinary families": "I came from nothing. Ordinary families have zero margin for error—one wrong major means four wasted years and hundreds of thousands of yuan in tuition."

But the skill file itself flags an internal tension among the "five inner contradictions": *voice of the poor vs. billionaire*. The model's value is simplifying complex competition into something actionable for information-poor families; its limitation is that it emphasizes filtering over creating—reducing life to not getting screened out.

Chapter 3: Five Mental Models

1. Social Sieve Theory. *Case:* choosing between a finance major at a non-elite university or materials science at a 985 school—Zhang recommends the latter, since finance depends heavily on credentials and connections, while a 985 brand passes résumé screens.

2. Choice > Effort. Effort in the wrong direction is waste. *Case:* strong programmers should skip grad school (tech hiring values skill), but mechanical/civil/biology/chemistry majors must pursue it—not for learning, but for credential laundering: "Know which sieve you're washing your degree for."

3. Employment-Based Backward Reasoning. Choose a major not by current interest but by what the industry looks like in four years. His journalism argument: median graduates end up in self-media or corporate PR, not state news outlets, because traditional media jobs are shrinking.

4. Class Realism. "Survive first, love later." *Case:* poor kids shouldn't study art—it costs hundreds of thousands extra and 80% of art graduates leave the field. "For rich kids it's pursuing a dream; for poor kids it's gambling." Not denying poor kids dreams—just the sequence: stabilize first, then climb.

5. Controversy as Propagation. In the attention economy, being discussed beats being liked. The journalism remark triggered media coverage, professor rebuttals, and trending topics—while Fengxue Weiwei's orders soared. A double-edged sword, but overall effective.

Chapter 4: Expression DNA

The skill also encodes *how* Zhang spoke:

Syntax: short sentences in rapid rhythm, punctuated by quotable punchlines.
Vocabulary: "finding a rice bowl" instead of career planning; "don't get screened out" instead of educational equity.
Northeastern dialect: words like *gabacui* (crisp/decisive), *zheng mingbai* (figure it out).
Rhythm: setup → subversion with facts → punchline → repetition.
Humor: hyperbole pushed to absurdity ("knock the kid out before letting them study journalism").

Chapter 5: The Nüwa.skill Technology

Nüwa.skill distills a person's thinking into a runnable AI skill via a six-dimension research process:

| Dimension | Zhang Xuefeng application | |---|---| | Books & systematic thought | 5 published books distilled | | Interviews & talks | 15+ in-depth interviews analyzed | | Expression style DNA | Short-sentence rhythm, dialect, hyperbolic humor | | Third-party views & criticism | Analyses from TMTPost, Huxiu, Sanlian, etc. | | Major decision analysis | 11 decisions, from tutor to founder | | Complete life timeline | 1984 birth → 2026 death |

Workflow: input a name → six parallel research agents → cross-validated mental model extraction → structured SKILL.md construction → quality verification (3 known test cases + 1 edge case + style testing).

Other distilled figures include Steve Jobs (minimalism, reality distortion field), Elon Musk (first-principles thinking), Naval Ravikant (leverage), Charlie Munger (latticework of mental models), Richard Feynman (learning by teaching), and Nassim Taleb (antifragility, barbell strategy).

Chapter 6: When Thinking Can Be Installed

The project raises genuine ethical questions. Released weeks after Zhang's death, it inevitably rides on the attention his passing generated—yet it is not mockery but rigorous study of his cognition, open-source and free. Given Zhang's lifelong pragmatism, a project extending his usefulness to more people might be something he'd endorse.

Practically, it democratizes advice once available only through crowded livestreams: anyone can now "consult Zhang Xuefeng" anytime.

It also sharpens a larger question about AI impersonation: this is not an AI *quoting* Zhang but an AI *thinking like* him—the former is entertainment, the latter a tool. But as AI gets better at mimicking human thought, the line between tool and replacement blurs. When a person's thinking can be encoded, installed, and infinitely copied, does their uniqueness still matter?

Conclusion

> "Life is fun. I'd come back for another round."

Zhang Xuefeng's arc—from poor rural kid, to bottom-rung Beijing salesperson, to 40-million-follower educator and billionaire founder—is a classic underdog story he knew required luck as much as effort.

zhangxuefeng-skill is, in a sense, a continuation of that pragmatism: converting a lifetime of accumulated judgment into code that can be inherited and used. Not resurrection, not replacement—a new form of remembrance: not remembering what he said, but how he thought.

When a mind can be installed, its influence outlives the body that produced it.

References

1. Huashu. zhangxuefeng-skill [open-source project]. GitHub. https://github.com/alchaincyf/zhangxuefeng-skill 2. Zhang Xuefeng. *You Are One Book Away From Grad School Success*. Beijing United Publishing, 2016. 3. Zhang Xuefeng. *Choice Matters More Than Effort*. Beijing United Publishing, 2021. 4. TMTPost. *Zhang Xuefeng: The Era's Most Complex Educational Symbol*. 2024. 5. Sina Finance. *Deng Qingxu In-Depth Dialogue with Zhang Xuefeng*. 2024.

*Note: Some details (e.g., the 2026 death date, company revenue) reflect the original post's claims and cited sources; readers interested in technical specifics should consult the GitHub repository directly.*