Chaos engineering sounds like something only hyperscalers with armies of SREs can do safely. Infrastructure-layer tools kill servers, VMs, pods, and inject faults - then you learn about uncovered failure modes and how to improve and operate better when failures actually happen. However, the infrastructure-level approach looks daunting for smaller teams. It is also context-blind - you can't scope the blast to 1% of internal test users.

This talk introduces you to the idea of controlled microchaos: runtime-controlled fault injection via feature flags that's targeted, instant, reversible, and SLO-safe. It's the on-ramp that lets a handful of people build resilience and confidence without betting production workloads. You'll leave with a concrete first experiment you can run next week - and the properties that make it survivable.

Category: Infrastructure / Platform Engineering
Level: 200 (intermediate)
Audience: Developers; Architects / Technical Team Leads / Product Managers; Infrastructure / DevOps / SysOps / Platform Engineers