Execution Trees, Not More Logs: A Better Debugging Model for AI Agents

Execution trees make the structure and causal relationships of AI agent runs explicit, revealing nesting, fallbacks, retries, and parallelism, and enable determ