Elizabeth at SigNoz asks what happens to site reliability engineers, who keep services running, when AI takes over routine incident response. The harder failures still need human judgment, but the everyday work that builds that judgment may disappear.
- Experienced engineers lose practice. Reading raw logs and testing explanations builds a mental model of the system. Accepting an AI diagnosis can bypass that work.
- New engineers may never build the skill. Automating the introductory tasks also removes opportunities to learn from manageable failures.
- Keep people actively involved. Elizabeth proposes deliberate human-practice opportunities and investigation alongside AI, beyond approving its recommendations.
Caveat: this is a vendor-affiliated argument drawing on automation research and examples from other professions, not a measured study of SRE deskilling. Its 95% automation scenario is a framing question, not a demonstrated adoption rate.
The 118-comment thread on Hacker News adds a practical disagreement: which activities build essential understanding, and which should disappear as systems improve?
What the thread adds
- kccqzy — challenges the premise that recurring incidents should supply training: “Instead of using AI to assist with grunt work, use AI or human thinking to build more reliable systems with fewer incidents.”
- a012 — reports using Claude for routine signal correlation, but says making it useful requires substantial company-specific context. That is one practitioner’s account, not a comparative test.
- recroad — suggests scheduled programming practice and having the agent quiz the developer before a merge. These are proposed countermeasures, rather than demonstrated fixes.
- MetaMalone — offers a different interpretation of the change: “I suspect we are just replacing old skills with new skills.”
HN handles are pseudonymous and no per-comment scores are published. Ordering is HN’s own ranking; this is a slice of the thread, not a consensus.