Not looking for model recommendations. Genuinely curious what's failing for people running agents in anything serious. I'll go first. My agents work great until they don't. The frustrating part isn't when they fail obviously, it's when they fail silently. Wrong answer delivered confidently. Tool cal
