How software engineers at Silicon Valley startups spend their days has changed dramatically over the past few years.(硅谷初创公司的软件工程师的日常工作方式在过去几年发生了巨变。) The output is still the same: reams(大量（纸张、文字等）) of code. But instead of(而不是，代替) sitting and typing, they whisper into microphones. Their instructions are then transcribe(转写，转录)d using artificial intelligence and fed to AI agents(智能体，代理)—bots that can interact with software and, in this case, write it.
过去几年，硅谷初创公司的软件工程师的日常工作方式发生了巨变。产出依然相同：大量代码。但他们不再坐着敲键盘，而是对着麦克风低语。他们的指令随后借助人工智能转写，并输入AI智能体——这些智能体能够与软件交互，在此场景下，即编写代码。
Other white-collar(白领的，脑力劳动的) workers are not yet “in this world”, admits Colin Jarvis of OpenAI, maker of ChatGPT. The future of his employer—and the AI boom at large(整个的，总体的)—hinges(取决于，关键在于) on that changing. To justify their colossal outlays on data centres, America’s cloud giants, and AI labs such as Anthropic and OpenAI, need adoption of the technology to continue spreading rapidly.(为了证明在数据中心上的巨额支出是合理的，美国的云巨头以及Anthropic和OpenAI等AI实验室需要这项技术的应用继续快速普及。) A back-of-the-envelope(粗略的，草草估算的) calculation by The Economist finds that covering the cost of the data-centre build-out(建设，扩建) will require total annual revenue from the use of AI to reach $2.5trn per year by the end of the decade, up from around $150bn now.
OpenAI（ChatGPT的开发商）的科林·贾维斯承认，其他白领工作者尚未“进入这个境界”。他雇主的未来——乃至整个AI热潮——都取决于这一状况的改变。为了证明在数据中心上的巨额支出是合理的，美国的云巨头以及Anthropic和OpenAI等AI实验室需要这项技术的应用继续快速普及。《经济学人》的一项粗略估算发现，要覆盖数据中心建设的成本，到本十年末，AI应用带来的年总收入需达到每年2.5万亿美元，而目前约为1500亿美元。
So far uptake(采用率， uptake) has been strongest by far(显然，……得多) among software developers. Four-fifths of programmers say they use an AI coding tool, according to Stack Overflow, an online forum. SemiAnalysis, a research firm, reckons that for Anthropic and OpenAI coding now accounts for over half their combined annual recurring revenue—a trendy measure of sales that extrapolates from the most recent month.(研究公司SemiAnalysis估计，对于Anthropic和OpenAI来说，编码现在占其合并年度经常性收入的一半以上——这是一种时髦的销售衡量指标，从最近一个月的数据进行外推。) In June 2025 the combined annual recurring revenue of Cognition, Cursor, Lovable and Replit, four AI coding startups, was roughly $800m. Today it stands at $6bn.
迄今为止，软件开发者中的采用率遥遥领先。根据在线论坛Stack Overflow的数据，五分之四的程序员表示他们使用AI编码工具。研究公司SemiAnalysis估计，对于Anthropic和OpenAI来说，编码现在占其合并年度经常性收入的一半以上——这是一种时髦的销售衡量指标，从最近一个月的数据进行外推。2025年6月，Cognition、Cursor、Lovable和Replit这四家AI编码初创公司的合并年度经常性收入约为8亿美元。如今这一数字已达60亿美元。
