The bringing of "agentic" into the mainstream this year was a little odd, crewAI was discovered by a broader group, but many people had quietly been playing wit

The bringing of "agentic" into the mainstream this year was a little odd, crewAI was discovered by a broader group, but many people had quietly been playing with it for a while. The fundamentals of how LLMs are different from what came before it are still being understood. People keep thinking it's like the world they've always known with software and discovering it's not. That could be an issue with LLMs, or how/what they are used for. Seeing different ways takes time. I feel like we're still missing some essential building blocks in widespread use to not only creating agents but making sure they stay stable and consistent across model changes.