科技
人工智能正在改变软件工程——一些开发者怀念曾经的奋斗时光。
中文译文
2002年,当杰玛·罗杰斯开始攻读计算机科学学位时,软件工程被视为一种具有发展前景且收入丰厚的职业。它吸引那些喜欢深入研究代码、排查错误以及从零开始构建程序的人。
总部位于英国的软件开发者罗杰斯告诉CNBC,这是一个发展速度始终很快的行业。但人工智能的出现进一步加快了这一进程,她所认识的每一位软件工程师的日常工作中现在都离不开人工智能工具。在2026年Stack Overflow的一项调查中,83%的软件开发者表示他们至少使用过一种人工智能工具。
由于人工智能编程工具能够帮助工程师以前所未有的速度编写、调试和测试代码,一些人开始怀念他们过去的工作。罗杰斯一直很喜欢通过将代码的各个部分组合起来找到解决方案所带来的成就感,尤其是当她花费数天甚至数周时间试图解决一个棘手的错误时。“过去,当我面临挑战并努力解决它时,那种快乐与趣味的感觉如今已有所消失,”她说。“现在有时候只需几条提示就能完成任务,这固然很好——但这只是些零星的快乐,而非充实的快乐体验。”
罗杰斯并非唯一一个怀念过去那种并不遥远的日子的人。软件工程师们纷纷在社交媒体上表达对过去工作方式的怀念,那时他们可以沉浸在代码中,连续数小时高效工作。
不过,人工智能工具确实很有用,也大大降低了入门门槛。“这已经变成一个行业,许多在过去10年、5年,甚至3年前可能无法取得成功的人,现在都能在这里有所作为,”罗杰斯说。她表示,工程师们传统上在工作中学到的核心技能也不应丧失。人工智能工具和互联网服务可能会意外出现故障,工程师们仍需掌握如何自行排查问题并找到解决方案。
“多年前我们所学的那些技能正在逐渐消失,因为工程师们不得不急于完成工作,”罗杰斯补充道。
评估人工智能代码
这份工作本身也在发生变化。人工智能决策公司Taktile的首席技术官詹姆斯·布罗德黑德告诉CNBC,过去可能需要三周时间才能制作出一个原型的工程师,现在只需一个下午就能制作出10个原型。他说,其中最困难的部分在于“要有判断力,能够分辨出哪些原型是合格的”。“这些[AI]智能体生成的代码量如此之大,以至于[工程师]们已经无法全部阅读了,”布罗德黑德说道。“过去,人们还能逐渐了解整个代码库的结构,但现在一切都变化得很快。”
AI技能差距
总部位于伦敦的AI培训公司Mindstone的首席执行官乔舒亚·沃勒告诉CNBC,他发现那些懂得如何有效运用AI的工作者与那些不懂的工作者之间的差距正在不断扩大。
他说,过去业内会提到“10倍效率工程师”,这类工程师的效率远高于同僚。如今,AI进一步拉大了这一差距,出现了“100倍效率工程师”。
沃勒认为,企业将越来越需要决定是帮助员工跟上AI的发展,还是优先考虑那些能最大程度利用AI的人。但他也理解为何有些工程师会难以适应。他说,那些曾经在工作中表现出色的人往往是那些热爱写作和编写代码的人,而这些人并不一定是目前的顶尖表现者。
“这种痛苦是非常真实的,”沃勒说道。“你热爱做的事情,公司在某种程度上不再愿意为此付费,因为它不再像过去那样为公司创造价值。”
监督机器
但并非所有工程师都认为远离手工编码是一件坏事。拥有20多年经验的软件工程师、总部位于伦敦的虚拟办公室公司Hoxton Mix的首席执行官克里斯·西斯告诉CNBC,人工智能让开发人员有更多时间去解决有趣的问题。他将人工智能视为一名有能力但仍需监督的初级工程师。“重要的是要知道该把哪些工作委派出去,以及何时需要人类的判断力,”西斯说道。“如果认为人工智能能做到一切正确,那未免太愚蠢了。我知道,能够找出人工智能出错的地方并加以证明,会让我感到极大的满足。”
即便如此,西斯仍担心,如果人工智能接管了工程师们传统上用来学习的常规编码和调试工作,那么下一代工程师将会面临怎样的处境。他说,这些任务“是人们吸取经验、积累经验的过程”,企业需要确保人工智能成为学习过程的一部分,而非捷径。
对罗杰斯而言,重要的是能够掌控自己使用这项技术的方式。作为一名自由职业的开发者,她可以利用人工智能来自动化处理那些她认为繁琐的项目部分,而将更有挑战性的问题解决工作留给自己。她认为,努力解决难题而非立即将其外包给技术,其中蕴含着重要的意义。她说,挫折与不适是人类经历中不可或缺的部分。
“作为人类,我们需要在某件事上获得成就感,而成就感的形成离不开挑战,”她说。“除非我们能在生活的诸多方面去应对这些挑战,否则就永远无法体会到成功带来的那种美好感受。”
原文
When Gemma Rodgers began her computer science degree in 2002, software engineering was seen as a future-proof and lucrative career. It appealed to people who enjoyed getting into the weeds of code, tracking down bugs and building programs from scratch.
It's an industry that has always moved fast, Rodgers, a software developer based in the U.K., told CNBC. But the arrival of artificial intelligence has accelerated the pace, with AI tools now part of the daily work of every software engineer she knows. In a 2026 Stack Overflow survey, 83% of software developer respondents said they use at least one AI tool.
As AI coding tools help engineers write, debug and test code faster than ever, some are mourning the job they used to have. Rodgers has always loved the sense of accomplishment that comes from piecing together different parts of code to find a solution, particularly when she's spent days, or even weeks, trying to fix a stubborn bug.
"The sense of joy and fun that I used to experience by having a challenge and then having to work that out, it's been removed somewhat," she said. "Now you can sometimes do it in a few prompts, and it's great — but it's a breadcrumb of joy, not a full meal of joy."
Rodgers is not alone in feeling nostalgic about the not-so-old days. Software engineers have taken to social media to describe missing their old way of working, when they could get into a flow state with code for hours at a time.
Still, AI tools can be useful and have lowered the barrier to entry significantly. "It's becoming an industry where a lot of people have been able to thrive who wouldn't necessarily have been able to thrive 10 years ago, five years ago, even three years ago," Rodgers said.
It's also important that the core skills engineers have traditionally learned on the job are not lost, she said. AI tools and internet services can go down unexpectedly, and engineers still need to know how to investigate problems and work through solutions themselves.
"The sorts of skills that we were taught years ago are kind of being lost because [engineers are] having to just churn work out," Rodgers added.
Judging AI code
The job itself is changing, too. James Broadhead, chief technology officer of AI decision-making company Taktile, told CNBC that engineers who might previously have spent three weeks building a prototype can now generate 10 in an afternoon. The hardest part of that, he said, is "having the judgment to decide which of those are any good."
"The [AI] agents are spitting out so much code that [engineers] haven't read it all anymore," Broadhead said. "It used to be you'd build up a sense of comfort and awareness of how the whole code base was structured, but now the whole thing changes all the time."
AI skill divide
Joshua Wöhle, CEO of London-based AI training company Mindstone, told CNBC he sees a growing divide between workers who know how to use AI effectively and those who don't.
The industry used to talk about "10X engineers," he said, who were dramatically more productive than their peers. Now, AI is widening that gap even further to produce "100X engineers."
Wöhle believes companies will increasingly have to decide whether to help employees catch up with AI or prioritize those who can get the most out of it. But he also understands why some engineers might struggle to adapt.
The people who once excelled at the job were often those who loved writing and working through code, he said, and those are not necessarily the current top performers.
"That's the grief that is very, very real," Wöhle said. "The thing you love doing is no longer the thing the company should be paying for, to a degree, because it's no longer adding the value to the company that it used to."
Supervising the machine
But not every engineer sees the move away from hands-on coding as a bad thing. Chris Sees, a software engineer of more than 20 years and CEO of London-based virtual office company Hoxton Mix, told CNBC that AI gives developers more time to tackle interesting problems. He views AI as a capable junior engineer that still needs supervision.
"The important thing is knowing what to delegate and when human judgement is needed," Sees said. "We'd be silly to think AI gets everything right, and I know I get a lot of satisfaction out of proving and spotting where it has gone wrong."
Even so, Sees worries about what happens to the next generation of engineers if AI takes over the routine coding and debugging through which they traditionally learned. Those tasks are "where you learn lessons and build scar tissue," he said, and companies need to make sure AI becomes part of the learning process, rather than a shortcut.
For Rodgers, what matters is having control over how she uses the technology. As a self-employed developer, she can use AI to automate the parts of a project she finds tedious, while keeping the more interesting problem-solving for herself.
She believes there's something important about working through a difficult problem rather than immediately outsourcing it to technology. Friction and discomfort, she said, are vital parts of the human experience.
"As humans, we need to feel accomplished in something and part of accomplishment requires challenge," she said. "Unless we are able to work that through in so many different aspects of life, then you never get the full force of the good feeling on the other side."
