A collection of browser-native, interactive labs covering the foundations of decision-making under uncertainty — from Markov Decision Processes all the way up to Deep Reinforcement Learning. Every lab runs entirely client-side; no install, no servers.
Build intuition for states, actions, transitions and rewards. Value iteration, policy iteration, and the Bellman equations — all visualised on an interactive 3D gridworld.
Open lab → 02 · Deep RLFrom the Q-table to the Q-network. Build a vanilla DQN, then layer in Prioritised Replay, Double targets, and the Dueling architecture — every concept visualised end-to-end.
Open lab → 03 · PlanningThe algorithm behind AlphaGo. Selection, Expansion, Simulation, Backpropagation — see how UCB1 builds an asymmetric tree, then play Tic-Tac-Toe against a live MCTS agent.
Open lab → 04 · Policy OptimisationThe natural sequel to Deep RL. Why DQN can't drive a car, why REINFORCE is so noisy, how Actor-Critic, A2C, A3C, PPO, DDPG and SAC each fix the next problem. Two playable envs included.
Open lab →Who deserves what when players cooperate? The Shapley value — the only fair split satisfying four natural axioms. Coalition-building, voting power, UN Security Council, airport cost-sharing, and SHAP for ML interpretability.
Open lab → 06 · Mechanism DesignSame bidders, different rules — different winners and prices. English, Dutch, First-Price, Vickrey, All-Pay, VCG. The Revenue Equivalence Theorem with a live Monte-Carlo. The mechanism behind Google Ads.
Open lab → 07 · Social ChoiceOne election, many winners. Edit a preference profile and watch Plurality, Borda, STV, Condorcet and Kemeny each pick a different champion — then meet Arrow's impossibility theorem and where voting powers AI.
Open lab →