When it comes to the general public, it requires some work that will get alot of attention. See for instance deepseek. Some of the my "normy" friends are even aware of the company despite not being into ML.

Maybe something regarding the foundational stuff regarding the formulation of Quantum Mechanics or Quantum Information Theory

In the area of deep learning based simulations, one good example of an open software is netket. The researcher their is pretty active in terms of github/gitlab/huggingface ecosystem.

The way I do it is to create exercise where I am required to fill as much of details as possible in the must explicit way. Sometimes, this is in the form of lecturing my friends for several hours. Or reproducing some computational results from computational physics (writing software and getting results correctly requires alot of details getting right). Similarly, when I studied such notes (for some of which I did not have accompanying course), I tried to fill as much of details as possible in the notes and do the exercises.

The ultimate state you want to reach is one where you can derive some results topic with enough details from scratch (without looking at the notes), like the way one can remember movie plot or something.

If you can take this process of understanding (and getting details right) very seriously + feynman technique, these can be helpful. But, you need to be really serious about learning and applying them (e.g. reproducing some results)

Yes, of course he was well-trained and had the enough background, but also the problem at the time was the type of problem that was solvable (i.e. no limitation in terms of tech) and that required new framework with new understanding.