Where David Silver's lectures teach you why reinforcement learning works, this course gets you training agents that actually do, and the two together are close to an ideal pairing. Hugging Face built it the way they build most of their learning material, code first and friendly, so instead of pages of notation you open a notebook in the first unit and train an agent, watch it fail, tune it, and watch it improve. It walks through the modern toolkit rather than reimplementing everything from scratch, so you work with libraries like Stable-Baselines3 and CleanRL that are genuinely used in practice, across a progression of environments from simple control tasks up to game playing, and crucially it covers the deep reinforcement learning era that the older lecture series barely touch. The Hub integration is a nice motivator, you push your trained agents, see them on a leaderboard, and get the small but real satisfaction of comparing against other learners, which keeps a self-paced course from going stale.
All of that makes a subject with a fearsome reputation feel genuinely approachable. The tradeoff is depth of understanding. This is deliberately practical, so the theory underneath is kept light, and you can finish a unit able to run an algorithm without being able to derive why it converges, which is exactly why I would not run it on its own. Reinforcement learning is also famously temperamental, training is sensitive to hyperparameters and random seeds in ways that supervised learning is not, so you will hit runs that quietly fail or refuse to learn, and the course cannot always tell you why, which can be frustrating without a theory foundation to fall back on.
And being self-paced with no instructor, untangling a stuck agent is your problem to solve. Take it for what it is, the hands on half of a reinforcement learning education, run it alongside or after a theory course rather than instead of one, and it is the best free on ramp into deep RL that I have come across.