HERA 2.0: Human-Governed Agentic Software Development with External Cognitive Orchestration -- An Industrial Experience Report on AI-DLC with ChatGPT Web as Cognitive Continuity Layer

Authors

• Manuel-Enrique Puebla-Martinez, Universidad del Trabajo del Uruguay, Montevideo, Uruguay

Abstract

The adoption of LLM-based agents in industrial software engineering has surfaced a structural problem: local agents can conduct AI-DLC workflows correctly while simultaneously pursuing incomplete or misguided problem interpretations when context is fragmented or prompts are imprecise. This experience report presents HERA 2.0, a methodology that addresses this problem through architectural separation of responsibilities: local agents perform situated execution, while ChatGPT Web acts as an external cognitive continuity layer. Evidence classification in HERA uses four operational states that naturally corresponds to the (T, I, F) structure of neutrosophic evidence degrees. An anonymized industrial case on POS-switch integration illustrates differences between HERA and direct AI-DLC.

Recommended Citation

Puebla-Martinez, Manuel-Enrique. "HERA 2.0: Human-Governed Agentic Software Development with External Cognitive Orchestration -- An Industrial Experience Report on AI-DLC with ChatGPT Web as Cognitive Continuity Layer." Neutrosophic Sets and Systems 101, 1 (2026). https://digitalrepository.unm.edu/nss_journal/vol101/iss1/14