You can play it here.
This was implemented entirely by prompting a local LLM (Qwen2.5 Coder 32B Instruct MLX 4-bit) running on my M4 Max 48GB Macbook Pro. Total time was maybe an hour. This implementation only runs a single level, has a fixed screen size, and other limtations
Maybe none? The first prompt pretty much generated something usable. That could definitely be due to it being such a popular example project in the past. However, it was pretty complete right away.