El Agente Potente
Purpose and distinctive contribution
Automate high-throughput atomistic simulations using machine-learning interatomic potentials.
Combines typed, provenance-aware execution graphs with a complementary coding mode; LLM planning is separated from deterministic scientific calculation and validation.
This page describes the specific system in the cited study, not every capability of the wider El Agente platform.
Publication and full author credit
Tsz Wai Ko; Jiaru Bai; Thomas Swanick; Yeonghun Kang; Changhyeok Choi; Angelina Qihong Jiang; Aiwei Yin; Varinia Bernales; Alán Aspuru-Guzik. (2026). El Agente Potente: High-Throughput Agentic Atomistic Simulations. arXiv:2609.14840.
Publication or posting date: . Preprint; journal publication not independently verified.
Research roles and contributions
Prof. Varinia Bernales and Prof. Alán Aspuru-Guzik are credited as coauthors alongside the complete team above. Detailed conception, supervision, implementation, and maintainer roles have not been independently verified for this page and are not inferred from author order.
Software, artifacts, and limitations
Paper-linked research artifacts → Paper-linked user cases; not verified as complete agent source. Availability evidence in the manuscript.
Capabilities are bounded by the tasks, tools, models, and evaluation conditions reported in the study. No universal speedup, reliability, or fully autonomous experimental capability is claimed here. Consult the linked manuscript’s methods and limitations before applying its results to a new workflow.