Or the training pushes it into the "Google it yourself" annoyed forum user mode. Maybe that points out wrong assumptions. Maybe it hallucinates that the assumptions are wrong. That is IMO more annoying than the sycophantic one.

As OP says, this is probably a by-product of them trying to "fix" the problem where the user can question a correct answer and it starts to sycophantically correct itself.

Just looked it up and there is IRCv3 to fix this, but idk what the state of that is.

Idk what exactly happend. But somehow they got approval from court to collect evidence of the crime (electronic devices) via house search. As said, the harm caused by the house search was later deemed unproportional to the crime.

My guess is that often reviews are generalizing (easy mistake to make). E.g. they say "service is slow", when they should say: "When I was there on Thursday at noon service was slow for my table".

For example NPCs in games that have complexity that previously was not possible.

Good games often push the boundaries a bit, so should be a good example.

Of course now we can start arguing that there isn't a lot of investment into gaming currently, because it all goes into AI. Too bad.

Even SpaceX is not profitable because of Starship.

Didn't make the problem go away.

Of couse it can be argued that PHP can be used outside this one way as well, but even e.g. https://symfony.com/doc/current/mercure.html is using golang.

To me looks like we aren't and your personal projects don't support the opposite.

So given there are constraints, resources must be rationed/allocated somehow.

Maybe we now have the technologies to remove this resource constraint. We'll see.

It is physically possible, but it won't have positive ROI so it is not viable.

If you have a paper/post doing the calculations for positive ROI, I'd be all ears. It can even have the optimistic Elon assumptions about price of mass to orbit.