The low-order bits can have huge impact. This is what Nicholas Nassim Taleb calls "fat-tailed processes": processes which rarely yield occurrences outside nominal range ("tail"), but such occurrences have the highest impact ("fat"). This is the case for anything related to safety: seatbelt, smoke detectors, guardrail, vaccines, etc. And one could formulate ridiculous umeshisms when talking about fat-tailed processes: if you've never died in your car, it means you're caring too much about your seatbelt; if you never suffered from 3rd degree burns, it means you invested too much in smoke detectors; if you've never had smallbox, if means you waste too much time getting vaccinated; if you've never murdered someone, it means you're not having enough fun with knives; etc.

So the general principle should be clarified as: _concentrate on the high-order bits, *iff the low-order bits don't matter*_. This is arguably the case for the examples given by the blog post, as missing a flight isn't really the end of the world. Still, it's a huge waste of time _on a single trip_, so you benefit from not caring about it only if you travel often enough. This shows than even in case where low-order bits aren't dramatic, they still need to be taken into account into a cost-benefit analysis.

Contrarily to carbon byproduct, we know how to confine fission byproduct, at least short term (which is still better than neither short term nor long term).

They're designed to sustain an airliner crashing into it so I guess bomb or missile is okay, although that's another story if that would happen from inside the reactor.

Interesting, but I doubt these are a majority.

I also don't think a plane is comparable to warfare in this aspect, a falling plane is an accident (or a relatively cheap act of terror), a missile can be a ballistic one, I don't think you can protect anything above ground from targeted destruction intent

Their users growth has been lowering towards 0 for a few years and it's expected that users will eventually shrink. It's just symbolic when you pass the tipping point (although this might be local and not be the definitive tipping point).

This isn't atypical for high performance network software, AFAIK. Single-thread with aio.

Because when people try to over-engineer something as simple at writing notes, they can loop on it forever.

He has had Neil DeGrasse Tyson and Brian Cox, for what it's worth.

The ryhtm should accelerate each N clicks so that the game eventually ends. Otherwise no matter how many points you get by clicking fast it's possible to accumulate the same amount by playing it safe.

It does. If you use the "autoplay" code elsewhere, it definitely starts speeding up.

That's the first interview. They usually test your ability to design systems and stuff in a second interview.

Labor is not productivity. Labor is labor, and if badly aimed, might as well be completely useless. Upper management has the power and responsability to make decisions which can have colossal implications on the overall productivty of the company towards its goal of selling what it sells. The real issue though, is that even though labor can approximately be measured, the impact of decisions not so much; whatever decision is taken becomes the "new normal" and it's hard to know how much it was a productive or improductive one. I'm not really knowledgeable on how upper managers are evaluated.