AI agents are overhyped and most of them are simply not ready for mission-critical work.

However, the underlying models and architectures continue to advance quickly, and we can expect to see more successful real-world applications.