Los asistentes de programación suelen actuar como un cocinero que, para hervir un solo huevo, relee el libro de cocina entero. Al recibir la tarea de hacer una pequeña corrección, vuelven a abrir decenas de archivos y verifican vínculos ya conocidos. Esto es trabajo innecesario: energía que se desperdicia.
Los investigadores propusieron el método E3: primero evaluar la complejidad real, realizar el mínimo de acciones y ampliar la búsqueda solo si falla. Este algoritmo recuerda al cocinero experimentado que prepara un huevo con una receta probada, sin movimientos extra. Es un enfoque razonable en lugar de pánico.
El objetivo no es probar un robot, sino una IA con base ingenieril que no infle la tarea como un agujero negro que devora todo a su alrededor.
tal como lo escribieron los autores, en inglés arXiv:2607.13034
Large language model (LLM) agents increasingly automate multi-step engineering and informatics workflows, yet they rarely ask how much effort a task actually requires. They often follow a maximum-context-first strategy--re-reading files and dependencies they have already seen--turning a one-line edit into a small code-base audit. We argue the missing capability is task-aware execution-scope estimation: judging a task's difficulty, the information it truly needs, and the shortest reliable path before committing budget. We formalize minimum-sufficient execution and the Agent Cognitive Redundancy Ratio (ACRR), and propose E3 (Estimate, Execute, Expand): the agent estimates an initial operating point, executes a minimum viable path, and expands scope only when verification fails. On MSE-Bench--a deterministic benchmark of 121 edits in a capability-controlled simulator--E3 matches the strongest baseline's 100% success while cutting cost by 85%, tokens by 91%, and inspected files by 92%, and further beats a strong adaptive retrieval baseline by 16%; the gains survive held-out instruction wording and essentially every cost weighting. A companion real-model harness (LLM-Case) corroborates the effect on a live gpt-4o agent editing a real open-source library, with every candidate patch graded by actually running the project's real pytest suite against a measured oracle: the over-reading is milder but real, and E3 is the leanest and fastest policy at comparable task success--its one shortfall a provider rate-limit, not a wrong edit. We frame this as a controlled probe of execution redundancy, not a measurement of any deployed agent, and position task-aware execution as a step toward engineering-grounded AI (EGAI)--agents whose effort is anchored in the engineering reality of the task. We release the framework and benchmark.
Sin el método E3, el agente examinó 12 veces más archivos de los necesarios para una corrección trivial.