Tencent Executive: Most of Tencent's Code This Year is Generated by AI 腾讯高管：今年腾讯大部分代码都由AI生成

When Tencent Senior Vice President Dowson Tang announced that "most of the code this year was generated by AI," there was likely a tacit round of applause in the conference room. But how much of that applause reflected genuine excitement, and how much masked underlying unease? An era where engineers no longer write code sounds like a productivity revolution, but upon closer reflection, it feels like a quiet dissolution of identity.

"We've handed over the task of writing code to AI." This statement is made lightly, as if announcing an upgrade to the office printer. But programming has never been just about assembling code—it’s a dance of logic, a struggle of creators punctuated by sparks of inspiration. When this process is outsourced to a probabilistic model, are we surrendering mere repetitive labor, or the core vessel of creativity? "Regularly guiding and correcting what AI writes"—this new role sounds like that of a supervisor, but can an architect detached from the context of actual code truly guide, or would they merely be drawing in the sand? Architectural design, if not rooted in the messiness of implementation details, risks becoming a castle in the air.

As a tech giant, Tencent’s engineering culture undoubtedly has deep roots, but the scale of "most of the code" strikes at the self-perception of the entire industry. If even leading companies are betting so aggressively on AI-driven programming, do smaller teams struggling to survive still have the confidence to maintain a "pure" human development team? The tide of technology waits for no one. This may accelerate the shift of programming from a "craft" to a "managerial function," but the growing pains in between will likely be borne by countless mid-level programmers. Their value will no longer be measured by the density of keystrokes, but by their ability to critically examine and correct the outputs of an opaque "black box." This demands higher wisdom, not faster typing.

Meanwhile, with companies like Qianxun Intelligence establishing robotics firms and Zhiwei Chuangxin working on automated chip design, all signs point to the same overwhelming trend: AI is evolving from an "auxiliary tool" to a "primary agent of production." Anthropic’s seemingly contradictory appeal for "all to stop AI research" sounds more like a shrill alarm—when even researchers fear the breakneck speed of their own products, are those of us downstream running too blindly? Monetization is seen as DeepSeek’s "coming-of-age ritual"—a harshly astute observation that highlights the essence of today’s AI race: after the festivities of free and open-source, commercial viability will ultimately test the true caliber of all technologies.

Thus, what Dowson Tang presented may not be a solution, but a sharper question. When code can be generated in bulk and robots begin to manufacture, how should we define and measure the core value of an engineer or a developer? Is it the moment of "creation" from nothing, or the supervisory ability to "guide" and "correct"? AI hasn’t made us lazier; it has simply forced us to confront a long-buried question: how many among us are merely using diligence to mask a poverty of thought?

The future where AI generates code has already knocked on the door ahead of schedule. Some are cheering as they push away their keyboards, ready to strategize; others stare at the smooth yet unfamiliar characters on their screens, feeling for the first time the visceral reality that their familiar battlefield is rapidly dissolving and being reconstructed. This is not prophecy—it is a fact unfolding before our eyes.

腾讯高级副总裁汤道生宣布“今年大部分代码由AI生成”时，会议室里大概响起了心照不宣的掌声。但这掌声里，有多少是真诚的振奋，有多少是掩饰的不安？一个工程师不再写代码的时代，听起来是生产力革命，细想起来却像一场静悄悄的身份消解。

“我们把写代码的工作都交给AI了。” 这句话说得轻松，仿佛是在宣布办公室的打印机升级了。但编程从来不只是代码的堆砌，它是逻辑的舞蹈，是创造者的挣扎与灵光乍现。当这个过程被外包给一个概率模型，我们交出去的，究竟是低级的重复劳动，还是创造力的核心载体？“定期指导、修正AI写的东西”——这个新角色听起来像监工，但一个脱离了具体代码语境的架构师，其指导会不会沦为在沙上画图？架构设计若不扎根于实现细节的泥泞，很容易变成空中楼阁。

腾讯作为巨头，其工程文化必然有深厚积淀，但“大部分代码”这个量级，冲击的是整个行业的自我认知。如果最头部的公司都如此激进地押注AI编程，那些挣扎在生存线上的中小团队，还有底气养一支“纯粹”的人类开发团队吗？技术浪潮从不等待。这或许会加速编程从一门“手艺”向“管理职能”的转变，但中间地带的阵痛，恐怕要由无数中层程序员来承受。他们的价值不再体现于敲击键盘的密度，而是面对一个不透明的“黑箱”产出，进行批判性审视和修正的能力。这要求的是更高的智慧，而非更快的手速。

与此同时，千寻智能成立机器人公司、智维创芯搞芯片设计自动化……所有的线索都指向同一股洪流：AI正从“辅助工具”升格为“生产主体”。Anthropic那句看似矛盾的“呼吁全员停止AI研究”，反而更像一声刺耳的警报——当连研究者都对自家产品的狂奔速度感到恐惧时，我们这些下游的应用者，是不是跑得太盲目了？收费被视作DeepSeek的“成人礼”，这话精辟得残忍，点明了当下AI竞赛的本质：在免费与开源的狂欢后，最终要用商业生存来检验所有技术的成色。

所以，汤道生展示的或许不是一个解决方案，而是一个更尖锐的问题。当代码能批量生成，当机器人开始制造，我们该如何定义和衡量一个工程师、一个研发者的核心价值？是那从无到有的“创造”瞬间，还是“指导”和“纠错”的监管能力？AI没有让我们变得更懒，它只是逼我们去面对那个一直被埋藏的问题：我们之中，有多少人只是在用勤奋掩盖思想的贫乏？

那个由AI生成代码的未来已经提前敲门。有人欢呼着将键盘推开，准备运筹帷幄；有人则盯着屏幕上流畅却陌生的字符，第一次真切地感到，自己熟悉的战场，正在快速地沙化、重构。这不是预言，这是正在发生的事实。