What changed
Kult, a London-based startup, announced it is using Google Cloud's Gemini models to power its gaming companion, which provides real-time assistance during gameplay. The system understands the game being played, the player's progress, and offers answers through voice and a second screen without interrupting immersion.
What this means in practice
Kult's solution addresses the challenge of providing timely, accurate, and cost-effective assistance during gameplay. The system uses a range of Gemini models, including the lightweight Gemini Flash tier for high-volume, repetitive tasks and more advanced models for complex reasoning. This allows Kult to handle thousands of interactions per game while maintaining speed and quality.
The companion provides answers in two parts: a voice response and a visual feed on a second screen. The visual feed includes images, embedded video, maps, and interactive tools tailored to the game.
Kult builds its knowledge base for each game automatically, ensuring it stays current with updates and patches. The system also adapts to individual players, learning from their interactions and tailoring responses to their preferences.
Limitations and availability
Kult is currently in alpha with a select group of players and plans to expand over the coming months. The solution is built using Google Cloud's Gemini platform and the Gemini Developer API, with backend operations running on Google Kubernetes Engine. The system's architecture allows for scalability and the addition of new visual answer types and devices without rebuilding the front end. Kult's integration with Google Cloud enables direct access to engineers when needed, supporting ongoing development and improvements.
The supplied announcement does not specify exact performance benchmarks or independent testing results. The system's effectiveness relies on the accuracy of the Gemini models and the quality of the knowledge base built for each game.
