Title: Cheap Fun: Accelerating Game Pre-Production via LLM-Based Vibe Coding The emergence of "Vibe Coding" promises a democratization of software development, empowering individuals with no programming knowledge to create functional software. This thesis investigates the feasibility and effectiveness of this approach specifically for rapid game prototyping using Large Language Models (LLMs). The study aims to quantify the "Quality Gap" between human-crafted and AI-generated games while identifying the optimal tools and prompting techniques for this new workflow. We established a rigorous methodology for performance analysis, generating three distinct game prototypes via LLMs to assess development efficiency, while selecting two for strictly comparative benchmarking against their human-authored counterparts.