Hacker News Reader

The most obvious instance of this is probably the Comet 1. It took about 3000 cycles (about half of them simulated in a tank) to reproduce the catastrophic failure.

But there is data to back that up on planes that are 'battle tested' you can compute failure rates over a ton of data. With something like this, a custom '1 off'... not unless you build a bunch of them and pilot them to depth (hopefully by remote) and wait for failures.

> No one knows how many cycles this sub was safe to perform

One less than the number that it did.

Only if you want 0 dives as a safety margin.

That is on-brand for the CEO here.

For this design that may not have been a bad choice.

"Sir, you should consider having a safety margin that's not a negative integer."

"Does it absolutely have to be an integer?"...

I'm fairly sure a safety factor capable of going NaN is not going to go well...