Tencent executive: Most of Tencent's code is generated by AI this year 腾讯高管：今年腾讯大部分代码都由AI生成

Tencent executives announced at a conference that this year, "most of the company's code is generated by AI." While this sounds impressive—like something out of a sci-fi film's futuristic factory—peeling back the dazzling technological veneer reveals what may be a quiet reshaping of programmers' professional identities, or even a glamorous prelude to "technological layoffs."

The executive's exact words were that engineers have "handed over the work of writing code to AI," shifting focus to "architectural design" and "regular guidance and corrections." This paints an optimistic picture of upgrade: from code laborers to AI mentors. But is reality really so seamless? "Most" is a vague term—is it 60% or 90%? What is the quality of the generated code? Are the engineers responsible for "guidance and corrections" seeing their workload increase or decrease? When AI can cover most coding tasks, the demand for junior programmers will inevitably drop sharply. This might not be "layoffs," but rather a gentle way of saying "we no longer need to hire as many people." The experience and growth path that newcomers once accumulated through writing large amounts of business code is being rapidly flattened by AI. Future engineers might indeed need to possess an "architect's mindset" from the start, but without solid coding practice, how can such a mindset be developed?

This isn't just an internal transformation within one tech giant. It's like a starting gun, signaling an irreversible inflection point in the software development paradigm. When one of the top tech companies embraces AI programming so aggressively, it means the entire industry's talent demand structure and skill evaluation standards are undergoing seismic shifts. Today you might say, "I can write code faster by hand," but tomorrow you may find that AI can generate a day's worth of work in a minute, with almost no syntax errors. The value of human programmers will increasingly concentrate in areas AI can't yet reach: profound insight into complex business scenarios, creative problem definition, trade-offs in system-level architecture, and most importantly—infusing AI-generated code with values and ethical judgment to ensure it doesn't spiral out of control or do harm.

Meanwhile, the industry itself is full of contradictions. Just as Tencent high-profile showcased AI productivity, another ironic headline emerged: AI company Anthropic called for "all to pause AI research." This is like a rocket pilot suddenly yelling "brake" while speeding forward—but the rocket's inertia may already make it unstoppable. This conflicted state of slamming on the gas while crying "danger" is the true snapshot of current AI development. The speed of technological advancement seems to have outpaced our ability to set guardrails and consider its consequences.

Other details are also thought-provoking. Discussions about when DeepSeek will start charging are merely the inevitable "rite of passage" for any successful tech product's commercialization—nothing to be alarmed about. Fei-Fei Li personally stepping in to "debunk" the world model rumors reveals the tension between concept hype and rigorous science in current AI research, where even top scholars must personally step in to clear the battlefield.

But perhaps the most human story among all this news is: Jensen Huang's reaction to ByteDance starting to use Arm CPUs—"I'm so sad." See, even the most prominent tech leaders show such human vulnerability and competitive instinct when faced with choices in business and technology paths. This reminds us that in this grand narrative driven by code and algorithms, what ultimately propels it forward is still human—human ambition, human fears, human choices, and the eternal competition and collaboration between people. AI is changing the way we create tools, but who defines those tools and for whom they serve—the baton remains firmly in human hands. It's just that the hand holding the baton needs to be more clear-headed and forceful than ever before.

腾讯的高管在大会上宣布，今年公司“大部分代码都由AI生成”。这句话听起来很酷，像科幻电影里的未来工厂，但扒开这层炫目的技术外衣，里面露出的，恐怕是程序员职业身份被悄然重塑，甚至是一次“技术性裁员”的华丽前奏。

高管的原话是，工程师们“把写代码的工作都交给AI了”，转而去做“架构设计”和“定期指导、修正”。这描绘了一幅美好的升级图景：从代码工人跃升为AI导师。但现实真的如此丝滑吗？“大部分”是个模糊的词，是60%还是90%？生成的代码质量几何？那些负责“指导、修正”的工程师，他们的工作量是增加了还是减少了？当AI能覆盖大部分编码工作时，团队对初级程序员的需求必然会锐减。这或许不是“裁员”，而是“不再需要招那么多人”的温和说法。那些本该由新人通过编写大量业务代码来积累的经验和成长路径，正在被AI迅速填平。未来的工程师，可能真的需要一入门就具备“架构师思维”，但没有扎实的编码实践，这种思维又从何谈起？

这不仅仅是一家巨头的内部变革。它像一声发令枪，宣告了软件开发范式一个不可逆的拐点。当最顶尖的科技公司之一都如此激进地拥抱AI编程，这意味着整个行业的人才需求结构、技能评价标准都在经历地震。今天你还能说“我手写代码更快”，明天可能就会发现，AI能在一分钟内生成你一天的工作量，而且几乎没有语法错误。人类程序员的价值，将越来越集中在那些AI暂时无法企及的领域：对复杂业务场景的深刻洞察、创造性的问题定义、系统级架构的权衡取舍，以及最重要的一点——为AI生成的代码注入价值观和伦理判断，确保它不会走向失控或作恶。

而此刻，行业本身也充满矛盾。就在腾讯高调展示AI生产力的同时，另一条新闻颇具讽刺意味：AI公司Anthropic呼吁“全员停止AI研究”。这像极了开着火箭冲刺的驾驶员突然喊“刹车”，但火箭本身的惯性可能已经停不下来了。这种一边猛踩油门、一边惊呼危险的拧巴状态，正是当下AI发展的真实写照。技术狂奔的速度，似乎已经超过了我们为其设定护栏和思考后果的能力。

其他细节也耐人寻味。DeepSeek被讨论何时收费，这不过是任何成功技术产品商业化必经的“成人礼”，没什么值得大惊小怪。李飞飞亲自下场“辟谣”世界模型，则揭示了当前AI研究中概念炒作与严谨科学之间的紧张关系，顶级学者也不得不亲自下场清理战场。

但所有这些新闻中，最有温度的一条或许是：黄仁勋得知字节开始使用Arm CPU后，表示“好难过”。看，即使是最顶尖的科技领袖，在商业和技术路径的选择面前，也会展露如此人性化的失落与竞争心态。这提醒我们，在这场由代码和算法驱动的宏大叙事中，驱动它前进的，终究还是人——人的野心、人的恐惧、人的选择，以及人与人之间永恒的竞争与协作。AI正在改变我们创造工具的方式，但工具最终由谁定义、为何服务，这根指挥棒，依然紧紧握在人的手中。只是，握着指挥棒的手，需要比以往任何时候都更加清醒和有力。