The Anxiety of the "Operator Immortal": In the Age of AI, Top Engineers' Skills Are Being Accelerated by Their Own Innovations
Hello everyone, I'm your financial journalist and economist.
Today, we're not talking about a dry technical report, but rather a "confession" that went viral within the tech community and then spread worldwide. The author is Liu Shengyu, a machine learning system engineer at DeepSeek and a top student from Peking University's Turing Class, revered by his peers as the "Operator Immortal."
The article, titled "I Have to Bury My Talent in Yesterday," focuses on one central issue: a top AI engineer is witnessing the rapid advancement of the very skills upon which his career depends, skills that are being developed—and even surpassed—by the AI systems he helped create.
This is not about spreading panic, nor is it a doomsday prophecy. Instead, it reflects the most genuine and painful reflections of a professional in the heart of the storm. Let me break down this issue in simple terms, covering five key aspects.
---
Who Is He, and Why Should His Anxiety Be Heeded?
First, let's understand who Liu Shengyu is. If we compare him to a factory, he's not the designer who draws the blueprints, nor the assembly line worker; he's the "tuner" who ensures the machines run at their highest speed and with the least energy consumption.
In the field of large AI models, algorithm scientists design the network architecture (how to make the AI smarter), while MLSys engineers like Liu Shengyu are responsible for integrating these complex systems into GPUs to make them perform efficiently. How challenging is this task? So challenging that there are no ready-made instructions; it all relies on experience and intuition. It's like an experienced traditional Chinese doctor feeling a patient's pulse or a skilled carpenter judging the best place to apply force when working with wood.
Liu Shengyu's resume is impressive:
- Top-tier education: Graduated from Peking University's Turing Class and led his team to win international supercomputing competitions.
- Rejected a Ph.D. from a prestigious university: He declined offers from Berkeley and Carnegie Mellon, preferring to directly solve practical problems and join DeepSeek.
- Core contributor: He recently contributed the main Attention operator to the DeepSeek V4.1 model, which significantly improved the model's efficiency.
Why Should His Anxiety Be Heeded? Because he's not just an outsider making random guesses. He's the one who built the tools. Now, he realizes that these tools are evolving into machines that can create their own successors. His perspective represents the real situation of the most critical, yet often overlooked, group in the AI industry.
---
What Is the "Technological Paradox"? The More You Work Hard, the Faster You Get Unemployed?
Liu Shengyu mentions a poignant concept in his article, which I call the "Paradox of Accelerated Self-Obsolescence":
1. The more perfect his code becomes, the faster and more cost-effective DeepSeek's models become.
2. The faster and more cost-effective the models are, the cheaper it becomes to train smarter AI programmers.
3. Once these AI programmers mature, their coding speed and efficiency surpass Liu Shengyu's.
4. As a result, every optimization he makes shortens the lifespan of his own skills.
It's like a master weaver: the faster and more exquisite his work is, the more it drives the development of automated weaving machines. When the machines are finally built, the weaver's skill is still needed (to design new patterns), but its value diminishes.
Liu Shengyu says, "AI can process 300 tokens in a second and write code in twenty seconds, while I can't." This isn't about who works harder, but about the difference in approach. Human engineers focus on optimization, while AI focuses on generation. When generation outpaces optimization, the traditional model of skilled craftsmen becomes obsolete.
---
From "Craftsman" to "Mechanical Pilot": A Fundamental Shift in Career Roles
Faced with this challenge, Liu Shengyu didn't choose to give up or remain blindly optimistic; instead, he proposed a practical solution: retraining for a new career.
- In the past (as a craftsman), his value lay in writing code by hand. He enjoyed the focus and satisfaction of his work. His core strengths were depth and experience.
- In the future (as a mechanical pilot), his value will lie in directing AI to write code. He won't be manually fixing every detail; instead, he'll set goals, monitor AI processes, and handle complex issues. His new strengths will be judgment, architectural vision, and the ability to manage AI.
Liu Shengyu uses a vivid metaphor: "I have to bury my talent in yesterday and become a mechanical pilot. I have more gears at my disposal, but I've lost the rhythm of creating things myself."
This change is profound. For most of us, it means that "being able to do something" is no longer the key; "being able to direct" becomes crucial. The future workplace will need fewer detail-oriented workers and more leaders who can define problems, evaluate outcomes, and coordinate resources.
---
The "Temperature Difference" in AI Anxiety Between China and the U.S.: Fear of Destruction or of Being Empty-Ceded?
This article gained international attention because it addresses two extreme views on AI:
- In the U.S., there's a fear of "survival": Voices like Anthropic CEO Dario Amodei, OpenAI's Sam Altman, and Elon Musk highlight the potential for AI to go out of control, threatening human existence through cyberattacks or biological threats. This is a "theological" fear concerning the fate of humanity.
- In China (and among people like Liu Shengyu), there's a fear of "ability erosion": People worry about their skills being diminished by AI. A survey of young people in both countries shows that while concerns about AI replacing jobs are declining, concerns about AI diminishing personal abilities are rising.
- The difference: Americans fear being destroyed, while Chinese (especially tech elites) fear losing their capabilities.
Liu Shengyu's article fills the gap between these two extremes. It depicts a "middle ground" where AI doesn't instantly lead to unemployment but rapidly renders existing skills obsolete. You must continuously evolve or risk becoming a mere operator or even redundant.
This anxiety is more realistic and relevant because it affects everyone's professional dignity and core competencies, not some distant apocalypse.
---
What Does This Mean for Us?
Liu Shengyu's story, though specific to top engineers, has broader implications for all of us:
1. Don't rely on skill barriers: Once, mastering a skill meant a lifetime of employment. Now, AI is lowering the barriers to these skills. The key is to "judge," "integrate," and "define problems."
2. Embrace collaboration with AI: You don't need to outperform AI; you need to understand the business, human behavior, and strategy. Learn to use AI as a tool, giving it commands and being responsible for the results.
3. Be wary of ability erosion: Relying entirely on AI can lead to the degradation of essential skills. Maintain some manual skills; even pilots need to know how to fly manually.
4. Find value in areas AI can't replace: Liu Shengyu turned to AI agents, which require creativity and judgment in complex situations, emotional understanding, and the ability to build new paradigms.
In summary, Liu Shengyu's article is a wake-up call. The AI revolution isn't about replacing humans with machines but about reshaping roles. For top talents like him, the challenge is to transition from craftsmen to leaders. For everyone else, it's about quickly learning to use AI tools, avoiding being replaced, and finding unique value in the new collaborative landscape.
AI is neither your enemy nor your nanny; it's a tool. Use it wisely, and you can achieve great things. Use it poorly, and it can crush you.
As Liu Shengyu says, "I certainly don't want to be revolutionized, but if it's inevitable, I hope it's me who does the revolution." This is the most sober and positive attitude towards the AI era.