There are also lock free algorithms written in Assembly, that doesn't mean the authors had any fun writing them.
Unfortunately Modern C++ is mostly found in conference talks.
Most the real life code looks quite Orthodox C++ to me, including from big name ISO members on their SDKs.
I would suggest that you might be projecting your experience on the rest of the industry. I’ve seen nothing but modern C++ at several companies for the last decade. I haven’t seen legacy C++ anywhere in a really long time, thankfully. I occasionally see very early C++11, which is definitionally modern though dated by modern standards. Baselines tend toward C++17 these days on average.
The magic is called productivity
that doesn't mean the authors had any fun
You went from implying that it can't be done, to saying a GC is about productivity, to saying it's about "fun". The goal posts are all over the place here.
Meanwhile, I've done it over and over in C++. The reality is that memory allocation shouldn't be a big part of a fast data structure anyway, let alone a good lock free algorithm, because if you're relying on allocation that much you are hopping around in memory, not to mention leaving a big part of your data structure to a black box.
Unfortunately Modern C++ is mostly found in conference talks. Most the real life code looks quite Orthodox C++ to me, including from big name ISO members on their SDKs.
I don't know what either of these is supposed to mean.
If you want to be explicit, tell me what algorithms need a garbage collector, because this all sounds a lot more like something you want to be true or want other people to agree with than something learned from extensive experience.