"I've failed 17 times with the exact same error": An agent debugging tale (blog.tonkotsu.ai)

After analyzing thousands of agent trajectories, we’ve seen all sorts of bizarre LLM quirks, this being one of them. Would love to hear other war stories also!