I Built TetrisBench, Where LLMs Compete at Playing Tetris. Here’s What I Found. | Andreessen Horowitz
Turning Tetris into a coding and optimization loop shows how GPT-5.2, Claude Opus 4.5, Gemini 3, Grok 4.1, and Sonnet 4 differ in long-horizon reasoning, strategy adaptation, intervention timing, and behavior under edge cases and shifting state.
I was playing Tetris99 one evening on my Nintendo Switch and started wondering whether I could build a version where I play against an LLM. Not in the sense of proving that a model could “beat” the game, but simply to see what it would feel like to play something familiar against a system that reasons very differently from a human. At the time, I wasn’t trying to create a benchmark. I didn’t have a leaderboard to climb, or a claim to make. I was mostly curious about the experience itself since Tetris is deceptively simple. At its core, it’s an optimization problem: every move is a tradeoff…
saved by
related reading
- How to Use LLM as a Judge (Without Getting Burned)manthanguptaa.in
- Patterns for Building LLM-based Systems & Productseugeneyan.com
- LLM Chess: Benchmarking Reasoning and Instruction-Following in LLMs through Chessarxiv.org
- combinatorial reasoning environments for LLMs and RLdjdumpling.github.io
- Frontier Coding Agents Can Now Implement an AlphaZero Self-Play Machine Learning Pipeline For Connect Four That Performs Comparably to an External Solver — LessWronglesswrong.com
- Learning to reason with LLMs | OpenAIopenai.com
- LLM Inference Performance Engineering: Best Practices | Databricks Blogdatabricks.com
- Noam Brown on X: "Implications of Large-Scale Test-Time Compute" / Xx.com
- Coding vs thinking — Paradigm 3paradigm3.org
- Playing chess with large language modelsnicholas.carlini.com
- Large Language Model: world models or surface statistics?thegradient.pub
- Can LLMs Be Computers?percepta.ai