What I Learned After a Year of Building Production-Grade AI Agents
Hallucinations aren't the biggest problem with LLMs. After a year building AI agents, we've found that the real challenge is the model's lack of common sense—and it shows up in surprising ways.