You learn fastest by doing. Ideally, build something that interests you. If you can't find anything, then pick literally anything. If you're at a total loss, do Project Euler problems in the language(s) you want to gain expertise in.

You can also always go back to your old curriculum and actually do it. It doesn't really matter if you can't grade it, just do the assignment and ensure you get the expected output. If you can't access your old curriculum for whatever reason, plenty of universities have open curriculum.

But really, I don't think the curriculum matters that much. Just build something you want to build, and don't use LLMs to do it. If you get lost, use Google, ignore the AI overview, and go to stack overflow, and try to really understand the answer.

I would say that you could try to understand the answer the LLM gives you as well, but I honestly think that'll be too much of a crutch for you at this point, and you should save that for much later in the process when you're feeling more confident in your abilities.

I wouldn't put yourself on a strict timeline, but you should be able to feel semi-confident within 2-6 months assuming you're doing a bit every day. If you're building something you are interested in, then the process will be quicker. You may not know the name of every single pattern, or the names of different architectures, but you should gain an inherent understanding of architectural design and good code vs bad code over time.

If it's a project you really enjoy writing, you'll write it once, and it'll be crap. You'll want to re-do it better this time, and you'll do it again, and it'll probably still be crap, but better. The third time rewriting it, you'll probably do it pretty decently.

The most important aspect to learning is failing, and not meeting expectations, whether they're your own or other people's. So go do something, suck at it, and try again with what you learned. You'll get there.