LLM Co-pilot for Utility AI: An LLM-based co-pilot is being developed to generate preliminary Utility AI setups from natural language descriptions, which users can then refine manually.

Data-Driven Utility AI Training: Research focuses on human-in-the-loop approaches for training Utility AI models from data, focusing on capturing the designer’s intent, explainability, and manual modifiability.

Grail Tool Plugin System: A plugin system is under development, allowing users to add custom configuration and debugging tools, extend the built-in debugger, and integrate with external visualization tools.

Expanding Game AI Templates: Work continues on AI templates for other genres, including turn-based tactical and real-time strategy games.