Anthropic初级岗锐减，AI代理工时超人类六倍
论证要点
- Anthropic内部研究显示，自2025年初以来，技术部门初级岗位占比从21%降至14%，新员工中初级占比从27%骤降至9%，2025年12月起初级职位空缺急剧减少。
- 作者将原因归于AI代理能力增长：到2026年8月，AI代理记录的工作时长是人类的六倍，写代码、找bug等任务越来越多交给Claude，且Claude处理最复杂工作的领域初级招聘下降最多。
- 生产力全面提升：工程师代码提交量增5.1倍，数据科学家查询量增3.1倍，法律团队合同处理量增3.1倍，销售成交增1.7倍，招聘和会计处理量各增1.4倍。
- 研究承认Anthropic是特例：极度依赖AI的软件公司，结果难以映射到整体经济，快速增长和自愿采用AI可能扭曲数据，更广泛经济尚未显示可测量的AI生产力增益。
- 北约特种作战大学研究员Nolan Lovett提出“认知公地悲剧”：企业用AI替代初级岗位获得效率收益，成本却由未来需要经验人才的社会承担，Anthropic符合这一模式。
- Lovett预测全面影响将在2030至2045年显现，因为现有专家是5至20年前培养的，软件工程、金融分析、法律研究是最暴露的领域；但该预测前提是未来人类仍从事这类工作，而Anthropic曾暗示程序员和呼叫中心工人最终需转向电工或护士等职业。
对实践的启示
对AI应用者而言，短期效率提升明显，但需警惕初级岗位消失导致未来人才断层；企业应平衡AI替代与人才培养，避免“认知公地悲剧”。
作者承认的反方与局限
研究承认Anthropic是特例，结果难以推广到整体经济；快速增长和自愿采用AI可能扭曲数据；更广泛经济尚未显示可测量的AI生产力增益。Lovett的预测也仅在未来人类仍从事这类工作的前提下成立。
原文摘引
By August 2026, those agents were logging six times as many work hours as human employees.
When individual companies replace entry-level positions with AI, they benefit from the efficiency gains while the costs get spread across everyone who needs experienced talent later.
So far, the wider economy hasn't shown measurable productivity gains from AI.
▸ 展开原文全文（12 段，抓取自 The Decoder）
Anthropic has nearly stopped hiring junior staff because AI agents now work six times as many hours
Anthropic is hiring fewer and fewer junior employees in technical roles, while the hours logged by its AI agents keep climbing.
According to an internal study published by the AI company, the share of junior positions in technical departments dropped from 21 percent to 14 percent since early 2025. New hires took a bigger hit. The junior share there fell from 27 percent to just 9 percent. Open junior roles declined especially sharply starting in December 2025.
Anthropic points to the growing capability of its AI agents as the reason. By August 2026, those agents were logging six times as many work hours as human employees. Tasks like writing code and finding bugs are increasingly handed off to Claude. Junior hiring declined the most in roles where Claude handles the most complex work.
Productivity rose across every department, according to Anthropic. Engineers shipped 5.1 times as many code commits, data scientists ran 3.1 times as many queries, and the legal team processed 3.1 times as many contracts. Sales closed 1.7 times as many deals, recruiters hired 1.4 times as many people, and accounting handled 1.4 times as many invoices.
The study itself admits that Anthropic is a special case, though. It's an extremely AI-heavy software company, and the results don't map easily onto the broader economy. Rapid business growth and voluntary adoption of AI tools could also be skewing the numbers. So far, the wider economy hasn't shown measurable productivity gains from AI.
Fewer junior roles now means fewer experts later
Researcher Nolan Lovett of the NATO Special Operations University recently described a "tragedy of the cognitive commons" in a research paper. When individual companies replace entry-level positions with AI, they benefit from the efficiency gains while the costs get spread across everyone who needs experienced talent later. Anthropic fits that pattern.
Skilled professionals only develop through years of practice in the very junior roles that are now disappearing. Lovett argues the full impact won't show up until 2030 to 2045 because today's available experts were trained 5 to 20 years ago. Software engineering, financial analysis, and legal research are the most exposed fields.
Lovett's prediction only holds if humans are still doing this kind of work in the future, though. That's something Anthropic itself probably doesn't expect: the company has suggested in the past that programmers and call center workers would eventually need to move into jobs like electrician or nurse.
AI News Without the Hype – Curated by Humans
Subscribe now