It's a bit like scripting in python vs C on limiting hardware. Even some people with fresh EE degree struggle with this. Modern design often heavily focuses on simulation and reusing the company's IP blocks. Companies try to push for this approach despite the inefficiencies exactly because they can then hire those fresh "unripe" engineers. Old-school engineers almost always come up with better ad-hoc circuit designs trough.
If you live near a university, you can probably sit in on lectures (analog-design, analog circuits, microelectronics). There are also many videos on YouTube. (If you're not really interested in repairs, skip those as that is slightly different skillset.) One really good channel is Sam Ben Yaakov's https://www.youtube.com/@sambenyaakov
Developing a mental model of circuits also takes practice. The practice can mean simulation or playing with an oscilloscope and components on a breadboard but you need the feedback loop: designing/tweaking the circuit, checking if the behavior matches your expectation, finding what you missed if there's something wrong. Start with simple circuits and gradually work your way up to more complex ones.
You're on a good track, didn't come to mind but prof. Razavi's lectures are great!
As the other person mentioned, you can do a lot in simulation. KiCad now has a builtin ngspice support (trough it's still a bit rough.) On windows, ltspice is another quite popular free option.
If you live near a university, you can probably sit in on lectures (analog-design, analog circuits, microelectronics). There are also many videos on YouTube. (If you're not really interested in repairs, skip those as that is slightly different skillset.) One really good channel is Sam Ben Yaakov's https://www.youtube.com/@sambenyaakov
Developing a mental model of circuits also takes practice. The practice can mean simulation or playing with an oscilloscope and components on a breadboard but you need the feedback loop: designing/tweaking the circuit, checking if the behavior matches your expectation, finding what you missed if there's something wrong. Start with simple circuits and gradually work your way up to more complex ones.