Designing, testing, and securing next-gen agentic AI systems

This cluster tracks the shift from simple code completion toward fully agentic workflows, exemplified by tools like Cursor where agent requests are rapidly gaining on traditional tab-complete usage. Practitioners are formalizing what “agents” are and how to structure them, while research like AgentDropoutV2 tackles information flow and pruning in multi-agent setups. New evaluation frameworks such as AI Gamestore explore scalable, game-based tests of general intelligence and agent behavior. At the same time, AI strategists are proposing foundational guidelines for safe deployment, underscoring that robust safety, governance, and evaluation must evolve alongside more capable agentic systems.