← All work

01 / The brief · Artificial intelligence coursework · CP468 · 2024

Your Move

A Python game-playing AI. Same decision, less computation.

Could our tic-tac-toe opponent think faster without choosing a worse move? Jaden and I tested that tradeoff for our AI class.

Play against the search
My role
Algorithm development + evaluationJaden and I extended TheSharperDev’s Python baseline with alpha-beta pruning and evaluated it together.
What we delivered
A Python opponent with alpha-beta search.Reported first decision: 26s → 2s.

Original course project · 2024

  • Python

Portfolio browser demo · 2026

  • TypeScript
  • React

02 / From problem to product

How we chose the solution

We compared minimax with alpha-beta pruning, skipping branches that cannot improve the decision. Less searching, without weaker play.

How we built it

Jaden and I extended the Python baseline. This TypeScript demo is a separate reconstruction using search, not a trained model.

Try thisPlay a move, then inspect the search. Compare minimax and alpha-beta: does the decision stay the same with fewer positions visited?

Explore inside the screen
Your Move · Browser reconstruction

Your move.

TypeScript demo · Play, then compare the search.

Your turn. You are X; AI is O.

Choose a square. Watch the search counts.

How the browser demo counts its work

Each count includes the starting board. A cutoff means the search skipped the remaining alternatives in a branch. AI turns run in a Web Worker; preset comparisons run in the page. Both explore only the 3×3 board.

03 / Findings + next step

What we found

Our report recorded 26s → 2s for the first decision, not rerun here. I learned to distinguish a faster AI from a better one, and evaluate both.

What I’d improve next

Next experiment: benchmark game-playing agents on larger boards, comparing search depth, runtime and win rate. Proposed, not built; no LLM agents in this project.

Next projectAfter the Shift