Brilliant is not an AI course platform in the way most of the entries on this site are, and that is exactly why I think it earns a place here. It teaches maths, science and computer science through short interactive lessons, and over the last couple of years it has built out a genuinely good set of AI material, including courses on how large language models work, an introduction to neural networks, and broader lessons on how the technology around us functions. The approach is the whole point. Instead of lecturing at you, it gives you visual puzzles and interactive widgets where you nudge a value and watch the result change, and through that you build real intuition for ideas like how a network learns from its errors or why attention matters in a language model.
I have used a lot of resources that explain neural networks, and very few leave you actually feeling like you understand them. Brilliant does. The format also fits real life, because the lessons are short enough to do over a coffee, they work well on a phone, and it is one of the few learning apps I have managed to turn into an actual daily habit. The catch, and you have to be honest with yourself about this, is that you will not write any real code here.
Brilliant builds understanding, not skills, so it will not make you job-ready, it will not give you a portfolio, and it is not a substitute for sitting down with PyTorch and a messy dataset. It is conceptual on purpose. The subscription also unlocks the entire library across every subject, which is great value if you are curious generally but more than you need if you only want the AI lessons, and the AI coverage itself is still fairly shallow compared with a dedicated machine learning course. So I would not recommend it as your only AI learning.
I would recommend it as the thing you run alongside a hands-on course, the part that makes the maths and the mechanics finally click while the other course teaches you to build. In that role it is close to the best money you can spend.