My Actual AI Stack for DevOps. What I Use Daily and What Is Overhyped.

Pure practice, zero sponsorship. This is the actual AI toolset I use every day to build enterprise infrastructure - with honest notes on what works, what's overhyped, and what I stopped using. Claude for architecture decisions. Cursor for daily coding. GitHub Copilot for autocomplete. And the one thing none of them do - own the outcome when things break. The honest before-and-after: before AI, writing a Terraform module took 45 minutes. Debugging a Kubernetes issue meant two hours of googling. Documentation was something I dreaded. After restructuring my workflow - Terraform module is 10 minutes of review instead of 45 minutes of writing. Debug sessions are 30 minutes of guided hypotheses. Docs get generated from code, I edit for clarity. The key insight, which took me a while to land on: AI replaces typing, not thinking. I still design the architecture. I still make the calls. I still take responsibility. The three-tool breakdown covered here - Claude for complex reasoning and architecture, Cursor for day-to-day Terraform and Bicep, Copilot for autocomplete and patterns I've written a hundred times - explains exactly what I reach for and when. Including what I stopped using…