Inspiration Strategy games are interesting not because of graphics or speed, but because of the thinking behind each move. Watching how different players approach the same problem often reveals more than the outcome itself. AI War Room was inspired by the idea of turning that experience into something observable by letting multiple AI agents compete while explaining their reasoning in real time. Instead of hiding intelligence behind a final result, we wanted to make decision making visible.
What it does AI War Room is a turn based strategy arena where multiple AI agents with distinct strategic styles compete inside classic logic and strategy games. Each agent analyzes the same game state, chooses an action, and explains why it made that decision. The user can advance turns, observe how strategies evolve, and see how different reasoning styles perform under uncertainty.
The focus is not on fast gameplay, but on understanding how intelligent agents think, adapt, and fail when faced with incomplete information.
How we built it The project was built as a web application using Lovable for rapid interface development and Google Gemini as the core reasoning engine. Each AI agent is powered by Gemini but constrained with a distinct strategic identity, such as conservative analysis, risk taking, or pattern seeking.
Game state is maintained in memory and passed into Gemini each turn so agents can reason based on the current board and past outcomes. The interface emphasizes clarity, showing the game state, the active agent, and the explanation behind each move.
Challenges we ran into One challenge was keeping the agents truly distinct. Without strong constraints, different AI strategies can converge on similar decisions. This required carefully designed prompts to enforce unique priorities and reasoning styles.
Another challenge was balancing game complexity with explainability. The game needed to be simple enough to follow while still allowing meaningful strategic differences to emerge.
Accomplishments that we're proud of We are proud of creating an interactive experience that feels engaging to watch and replay. Each agent consistently demonstrates a unique approach to the same problem, and the reasoning explanations make it easy to understand why certain strategies succeed or fail. The project successfully showcases multi agent reasoning using a single Gemini integration.
What we learned We learned that large language models can behave like distinct agents when given clear goals and constraints. We also learned that explainability is just as important as correctness when designing AI driven systems, especially in game and simulation settings.
What's next for AI WAR ROOM Future plans include adding more strategy games, allowing users to customize agent behaviors, and introducing difficulty scaling. We also want to explore visual summaries of strategy performance over time and give users more control over the simulation parameters.
Log in or sign up for Devpost to join the conversation.