AI handles incidents, engineers lose touch with their systems
Original: https://www.sylvainkalache.com/blog/ai-handles-incidents-engineers-lose-touch-with-their-systems
Source platform: blog
Author: Sylvain Kalache (Rootly)
Original date: 2026-09-04
summary_zh
Rootly 的 Sylvain Kalache（前 LinkedIn SRE）9 月 4 日发文，把当下 AI SRE / 事故响应工具的扩散放进 Lisanne Bainbridge 1983 年Ironies of Automation框架：自动化越成熟，人越没机会练手，罕见异常真正来时反而最差判断：AI SRE 普及后整体 MTTR 会下降，但复杂事故恢复时间会显著拉长Kalache 用航空业对照FAA 要求机长每 6 个月做含发动机失效等场景的复飞训练，原因是商用涡轮发动机每 10 万飞行小时不到一次 in-flight shutdown落到工程团队这边，混沌工程和桌面推演必须升级为"事故模拟器"他的提醒很直白：explanation and observation are not substitutes for practice
summary_en
Rootly's Sylvain Kalache places AI SRE's spread in Bainbridge's 1983 Ironies of Automation frame: the more mature automation becomes, the less humans practice; when rare anomalies arrive, responders are worst. He predicts aggregate MTTR falls while complex-incident recovery time rises; engineers must upgrade chaos engineering into incident simulators with pressured judgment against LLM stakeholders.
one_liner
AI 修事故越顺，工程师就越不会修事故；混沌工程必须升级成"事故模拟器"，让工程师在压力下做判断