@peterdrake @maxthefox @PTR_K @mattmcirvin @futurebird

And it will be almost certain death to set foot inside an airliner constructed by Boeing.

@PTR_K @nyrath @maxthefox @mattmcirvin @futurebird I like working in PRA space backstopped by design basis events. Define your frequency/consequence envelope, model your system, quantify your uncertanties, and interate the design to stay inside the envelope. Start with overconservative “bounding“ events then move on to the more time-consuming probabilistic analyses. This is generally only worth the effort in "failure is not an option" systems (ironically, you accept small scale failures as inevitable and build in resilience to reduce the chance of large scale failures...). The most critical piece is defining the outcomes you care about (are defending against). Set those without humanity and caritas and none if the analysis matters - who cares what the numbers look like if you're solving the wrong problem?

FMEA is appropriate for relatively simple systems with a small number of well defined failure modes. I've seen it applied to software and IMO it's worse than useless - it gives you a false sense of rigor, like a deterministic version of LLM slop.

[Citations on request for those compelled to dig into the gory details. The presence of @nyrath in the thread means there are plenty of fans of gory detail here but I don't want to hijack the thread anymore than I already have] #RiskAnalysisForTheOverlyCurious