Introducing Gemifish: A Revolutionary AI-Powered Chess Engine with Dynamic Personalities

In the world of chess AI, most programs tend to offer a straightforward, albeit sometimes predictable, gaming experience. However, recent advancements in artificial intelligence open new possibilities for more engaging and personalized interactions. Today, I am excited to introduce Gemifish, an innovative AI chess engine that combines the formidable strength of Stockfish with the versatility of large language models to deliver a uniquely customizable and entertaining chess experience.

What is Gemifish?

Gemifish is a browser-based chess engine that not only challenges players with a high-caliber opponent—ranking over 3,500 ELO—but also enables users to prompt it to adopt a wide range of personalities. Whether you want to face an “Aggressive Grandmaster” who sacrifices pieces with flair and taunts you on the way, or a “Playful Cat on Drugs” that makes hilarious, unpredictable moves, Gemifish makes it possible.

How Does It Work?

The core of Gemifish is a clever integration of two powerful technologies:

• Stockfish 17.1: Renowned as one of the strongest open-source chess engines, it is embedded directly in your browser to analyze moves and suggest optimal plays.

• Google Gemini (or similar LLMs): Utilized here to inject personality, humor, and style into the game. It acts as a “persona filter,” selecting moves based on the character prompt rather than solely on raw engine strength.

The process works as follows:

1. Stockfish calculates the top five best moves and assesses their strength.

2. Gemini receives this list and, considering the desired personality traits specified by the user, chooses a move that aligns with that character’s behavior.

Additionally, Gemini can provide in-character explanations for its choices, making each game not just challenging but also entertaining and immersive.

Achieving 3500+ ELO with a Language Model?

A common misconception is that large language models (LLMs) like Gemini are poor at playing chess due to their generalized training. In Gemifish, that’s intentionally avoided. The heavy lifting is done by Stockfish, which ensures the engine’s strength surpasses 3,500 ELO. The language model’s role is to select and interpret moves in a way that embodies the chosen personality—adding flavor rather than raw power.

Open Source & Accessibility

Gemifish is fully open source, enabling enthusiasts and