Thread by @NewAtlantisSun

Nuclear weapons are real, but what's more interesting is speculation about whether the actual ones we have still work. We have many empirical examples of extremely complex aerospace engineering lore being lost outright.

What tended to happen is that the absolute sharpest knife in the drawer would build something at just about the limits of his comprehension, because the budget was effectively unlimited as long as you could produce a set of capabilities. Then he retires.

His replacement is not as good. And there was a bunch of arcane process knowledge accumulated, much of it unique to specific production runs (cf the stories of chip fab yields cratering because someone changed their shampoo).

That process knowledge is really meta knowledge about the debugging process, which is notoriously difficult to systematize. It dies when you stop firing off production runs (ie, you stop building, remanufacturing, or even testing nukes or delivery vehicles).

There is no incentive to be the guy who says "I don't think this works". There is limited incentive to be the maintainer when you don't get to make cool new stuff (and selling ad software is more profitable), so there is a negative feedback loop until you hit median civil servant

Difficult to overestimate the technical fragility of systems way past their expected lifetimes which were initially designed at the bleeding edge of physically possible performance envelopes and subsequently maintained by people unable to demand replacement or adequate testing.

Aerospace lore we have lost:
"Fogbank", which was supposedly able to be reverse engineered and remanufactured a decade after it was discovered to be a problem en.m.wikipedia.org/wiki/Fogbank

B2 heat exchangers: greatgameindia.com/air-force-reve…

Much of the Saturn V project: google.com/amp/s/www.vice…

The F22 writ large: nationalinterest.org/blog/the-buzz/…