The bottleneck, and the way around it
You cannot expand working memory. It is a hard limit and no amount of training moves it. What you *can* do is change what counts as one item.
A beginner reading a chess position sees thirty-two pieces. A grandmaster sees five or six familiar structures. Same board, same working memory, wildly different capacity — because the grandmaster's units are bigger. This is the whole of expertise, in one sentence.
Why this explains a lot of your struggles
When a problem feels overwhelming, the usual reason is not that it is hard. It is that you have not yet chunked its components, so every sub-step is consuming a working-memory slot and there are none left over for the actual reasoning.
A student who has not automated basic differentiation cannot follow a derivation that uses it, because the differentiation is eating all the room. They will conclude they are bad at the derivation. They are actually bad at the prerequisite, and it is invisible to them.
Building chunks on purpose
- Automate the prerequisites to the point of boredom. If a sub-skill takes conscious effort, it is stealing capacity from the thing you're trying to learn. This is what drilling is actually for.
- Name the patterns. Giving a recurring structure a name is literally what makes it one item instead of five.
- Practise the components in isolation, then in combination. Both, in that order.
- Notice when you're overloaded and go back a level. Feeling lost in a lecture is usually a prerequisite problem wearing a disguise.