An early prototype for a map of knowledge: the concepts behind a skill, how they connect, and where you might go next. The first interface explores a small route through reinforcement learning.
Start with what you want to do
A course gives you a route. I’m interested in the map underneath it. What do you need to understand before a difficult idea starts to make sense? Which parts can you approach from another direction?
One small, inspectable map
The current interface contains six nodes around the goal “implement PPO”, a reinforcement-learning algorithm. Python and probability lead into Markov decision processes, with Bellman equations and policy gradients further along the map. Selecting a node reveals its description and a sample explanation of why it might matter next.
Understanding is more than completion
The prototype uses illustrative learner states. Real diagnostics, lessons, and assessment are planned next steps. The underlying design principle is to keep evidence of learning separate from an estimate of mastery: finishing an activity does not automatically mean you understand it.
Keep the first question small
An open map of knowledge is a large ambition. This first interface narrows the question to whether a learner can understand a suggested next step and see how it serves a goal. The larger graph, adaptive recommendations, and contribution model remain future work.