DeepSeek Engineer Says AI Could Soon Replace His Job, But He Still Feels Compelled to Build It
A DeepSeek engineer has offered a strikingly honest view of the future of artificial intelligence: the same technology he is helping to develop may soon be capable of replacing him.
Shengyu Liu, a kernel engineer at the Chinese AI lab DeepSeek who also writes under the name interestingLSY/intlsy, recently shared his thoughts in a long blog post about the accelerating pace of AI development, the future of programming jobs, and why he continues working on systems that could eventually outperform human developers.
His argument is not simply that AI is getting better. It is that AI is improving so quickly that the role of software engineers may change dramatically in the near future. Instead of developers manually writing and optimizing large amounts of code, the most valuable people may become those who know how to use AI tools effectively to build high-performance software at extraordinary speed.
Liu described how AI has evolved from a basic assistant that could check code, summarize documents, or answer simple questions into a far more capable system. Today’s advanced models can read complex software environments, interpret technical frameworks such as NVIDIA CUDA, and use tools to analyze and reason through difficult engineering problems.
For a kernel engineer, that progress is impossible to ignore.
Liu suggested that within six months to a year, AI systems may be able to produce programming work comparable to his own. He noted that AI can already process information at a speed no human can match, generating hundreds of tokens per second, issuing commands almost instantly, and producing usable code in a matter of seconds.
In his view, this creates an uncomfortable reality for developers. If AI can think faster, write faster, and iterate faster, then many traditional programming tasks may soon be handled more efficiently by machines.
Yet Liu says this does not make him want to stop building AI. In fact, he believes the opposite.
He explained that even if he personally slowed down, competitors would not. Other AI companies would continue pushing forward, and the technology would keep advancing regardless of his hesitation. From that perspective, refusing to work on AI would not prevent disruption; it would only mean giving up influence over how the technology develops.
There is also a personal motivation. Liu described the excitement of working on AI systems that exceed expectations and the satisfaction of contributing to rapid technological progress. But his reasoning goes beyond ambition. He believes the structure of the AI industry will determine whether the future is widely beneficial or deeply unequal.
The broader AI debate has grown more intense as leading figures in the industry continue to clash over safety, speed, and access. Anthropic CEO Dario Amodei recently called for a slowdown in AI development, prompting pushback from NVIDIA CEO Jensen Huang. Other major technology figures, including Elon Musk and OpenAI CEO Sam Altman, also weighed in on the discussion. Meta CEO Mark Zuckerberg has separately argued that his company has taken internal safety precautions without making dramatic public declarations about slowing model development.
Liu’s post enters that debate from a different angle. He is not only concerned about whether AI develops too quickly, but also about who controls it.
He warned of two possible futures. In one scenario, the world begins to resemble a dystopian Cyberpunk-style society, where the most powerful AI systems are controlled by a small number of companies or elites. In that world, only a limited group would have access to cutting-edge intelligence, while everyone else would be left with weaker tools and fewer opportunities.
In the other scenario, the benefits of artificial intelligence are distributed more broadly, with powerful AI made available openly and affordably to more people.
This is where Liu sees DeepSeek’s mission as important. He expressed doubt that companies such as OpenAI or Anthropic would make the most advanced AI freely and cheaply accessible to the public. In his view, DeepSeek is better aligned with the goal of open access, especially if frontier AI models become central to education, software development, research, business, and national competitiveness.
Liu’s position is controversial, especially because he frames access to artificial general intelligence as a matter of global power. He used a severe historical analogy to argue that allowing one group to control a transformative technology before others can catch up would be dangerous. His point was not subtle: if AGI becomes the most powerful tool in the world, then concentrating it in the hands of a few organizations could create enormous risks.
That belief appears to be the reason he continues working on AI even while acknowledging that it may threaten his own role.
For software developers, Liu’s comments capture a growing anxiety across the technology industry. AI coding tools are improving quickly, and companies are increasingly looking for ways to automate engineering workflows. While AI may not eliminate all programming jobs overnight, it is likely to change what employers value. Deep technical skill will still matter, but the ability to direct, evaluate, and improve AI-generated work may become just as important.
The future Liu describes is not one where humans disappear from software development completely. Instead, it is one where the fastest and most capable developers may be those who can combine human judgment with AI-powered execution. In that world, writing every line of code by hand may become less important than knowing what to build, how to guide the AI, and how to verify that the result is correct, efficient, and secure.
His message is both optimistic and unsettling. AI could unlock massive productivity, lower barriers to advanced software development, and make powerful tools available to people who previously lacked access. But it could also deepen inequality if the best systems remain locked behind a small number of companies.
For Liu, the answer is not to step away from AI development. It is to keep building, keep pushing, and try to ensure that the most powerful technology of the coming era is not reserved only for the few.
