02 · Systems study
AI Search Lab
A collection of small, inspectable implementations that make foundational AI ideas concrete—from graph search and minimax to logic and reinforcement learning.
Selected outcomes
- Compared BFS, DFS, UCS, and A* behavior
- Implemented adversarial search with minimax
- Explored logical inference and probabilistic reasoning
- Trained and evaluated a reinforcement-learning Nim agent
Why build the fundamentals?
Modern AI is often presented through polished APIs. Those tools are useful, but they can hide the decisions that make intelligent behavior possible: how a state is represented, which paths are explored, how uncertainty is handled, and what a reward function actually encourages.
This lab keeps the systems deliberately small enough to inspect. The goal is not production scale; it is building reliable intuition.
Experiments
- Measure how search strategy changes completeness, optimality, and memory use
- Stress-test admissible and consistent heuristics in A*
- Trace minimax decisions through small adversarial games
- Observe how exploration and reward design affect a Nim agent
Next iteration
The next version will turn individual scripts into reproducible notebooks with common evaluation helpers, diagrams, and short explanations of the result—not just whether an algorithm worked, but why its behavior changed.