• Sources: primary, discussion
  • Summary: The SigNoz newsletter post, dated 2026-02-28 and trending on Hacker News on 2026-10-30, argues that automating tier 1 and tier 2 incidents leaves SREs only the novel tier 3 failures while removing the practice that builds intuition for them. It cites deskilling findings from endoscopy and aviation and proposes deliberate human-practice incidents and human-in-the-loop designs over approval-only supervision.
  • Why it matters: Teams adopting AI SRE agents may need to schedule manual incident practice so on-call engineers keep the skills automation stops exercising.