这篇文章聚焦于重返职场的新手妈妈在软体开发领域面对的 AI 冲击。Danielle 在 2024 年中离开职场时，几乎没有人用 AI 写程式；一年后，AI 已成为预期标准，连原本强调工作稳定的程式设计也被「外包」给模型。Anthropic 与 OpenAI 于 2025 年 5 月推出编码自动化工具后，工作重心从真正编写程式转向监督 AI，且业界高层还预测 AI 会在未来 18 个月内撰写 Meta 大部分程式码。

对产假中的女性而言，问题不只是在技术上落后，而是职涯节奏被打乱后，还要追赶快速变化的技能门槛。英国一位专案经理表示，主管建议她在休假期间补 AI 能力，让她感到非常脆弱；她担心若不跟上，可能成为裁员目标。美国健康科技公司的资料工程师 Mary McCreary 则认为 AI 有助于理解同事的程式码，但也让她失去处理琐碎工作的时间，迫使她只面对更困难的问题；明尼苏达州另一位工程师则指出，AI 帮她在产后疲惫与脑雾下维持进度，2024 年 9 月初回归时公司还只把 AI 当作升级版 Stack Overflow，一年后却已要求所有改动先经 AI 检查，甚至用排行榜计算使用量。

求职市场同样被 AI 重塑，且对曾因生育中断工作的女性特别不利。Danielle 在生产前 3 个月遭裁员，之后投了 40 份申请，只有 1 份进入面试；她发现多数职缺都要求 AI 知识，却很少说明具体如何使用，让补强方向变得模糊。专家指出，职场往往把育儿离开视为「退出」而非「暂停」，而在软体业，AI 又加深了母亲与同事之间的落差，让原本就存在的不利条件被进一步放大。面对这种变化，有人考虑转向较不易自动化的职业，例如 Danielle 正考虑景观建筑；她甚至怀疑，在一个以 AI 生成与修补程式为主的未来，这样的工作是否还值得投入。

This article examines how new mothers returning to software development are meeting an AI-transformed workplace that feels unfamiliar and unstable. Danielle left work in mid-2024 when AI was rarely used to write code, but by the time she was ready to return a year later, AI had become the norm. After Anthropic and OpenAI released coding automation tools in May 2025, software work shifted from writing code to supervising models, and leaders such as Mark Zuckerberg and Sam Altman suggested that AI would take over an enormous share of coding, including most of Meta’s code within 18 months.

For women on maternity leave, the problem is not only technical obsolescence but also the loss of time during a period of rapid change. A UK project manager said her manager urged her to brush up on AI while away, which made her feel vulnerable; she did not want to spend her statutory maternity pay on an AI course and feared being vulnerable to layoffs if she fell behind. Mary McCreary, a US data engineer, found AI useful for explaining coworkers’ code, but said it removed the chance to do tedious work and left her with only hard problems. Another Minnesota engineer said AI helped her keep up despite postpartum fatigue and brain changes; when she returned in September 2024, AI was only a helper for troubleshooting, but a year later all code changes were being checked by AI, usage was tracked on a leaderboard, and by November 2025 she said Claude Opus 4.5 was so capable that she could do a quarter’s worth of a team’s work by herself.

The job market has also been reshaped, making re-entry harder for women returning after childbirth. Danielle was laid off three months before giving birth and later sent out 40 applications, but only one reached an interview. She found that most postings required some AI knowledge without explaining what that meant in practice, creating a stressful ambiguity. Experts said employers often treat maternity leave as an exit rather than a pause, and that AI has widened the literacy gap between mothers and colleagues while worsening inequality. Faced with these pressures, some women are reconsidering software altogether; Danielle is thinking about landscape architecture, and she says she does not find meaning in training AI or merely fixing AI-generated code.