Engineering systems are becoming increasingly data-rich, knowledge-intensive, and decision-critical. The emergence of ChatGPT and related large language models (LLMs) has created a new opportunity to connect human expertise, computational tools, and domain-specific reasoning in ways that were difficult to realize with conventional artificial intelligence (AI) alone. Yet engineering applications require more than fluent language generation: They demand reliable reasoning under constraints, adaptation to specialized terminology and data, transparent interaction with human experts, and practical integration with physical or cyber-physical systems. In this context, ChatGPT is moving from a general conversational assistant toward a broader engineering intelligence infrastructure.

We curated this special issue titled “Applications of ChatGPT” for these reasons. The issue brings together seven diverse and insightful contributions that demonstrate how ChatGPT-like models can be adapted, fine-tuned, grounded, and orchestrated for real engineering tasks. Together, these articles cover autonomous driving, reliability systems engineering, earth observation satellite scheduling, wind farm maintenance, cognitive diagnosis, smart firefighting, and industrial design automation. Through this collection, we seek to highlight a shared transition: LLMs are no longer used only to explain engineering knowledge but are increasingly being embedded into workflows as agents, decision-makers, knowledge constructors, and tool users.

In this special issue, two articles first provide broad perspectives on the engineering implications of LLMs. Zhu et al. review LLM-powered autonomous driving, analyzing how LLMs may support both modular and end-to-end driving systems by enhancing perception, prediction, planning, control, human-vehicle interaction, and reasoning over long-tail scenarios. Their review also raises important questions about safety, security, and whether LLM-based artificial general intelligence can contribute to high-level autonomous driving. Zhang et al. turn their attention to reliability systems engineering, using the reliability systems engineering (RSE) V-model to examine LLM applications across requirements, design, manufacturing, verification, and maintenance. Their perspective identifies opportunities for the use of LLMs in complex engineering life-cycle management, while emphasizing persistent challenges in domain data, system-level analysis, explainability, and output evaluation.

A centerpiece of this special issue is the work by Chen et al., who introduce AgentAD, an LLM-based multi-agent framework for autonomously designing algorithms for the Earth observation satellite scheduling problem (EOSSP). The EOSSP is a complex NP-hard optimization problem involving the selection, sequencing, and timing of observation tasks under limited satellite resources and diverse operational constraints. Traditionally, designing effective algorithms for such scenarios requires substantial expertise, repeated trial and error, and costly manual modification. AgentAD addresses this challenge by transforming natural language descriptions of EOSSP scenarios into executable algorithmic code. The framework organizes five LLM-based agents—Algorithm Designer, Algorithm Programmer, Algorithm Optimizer, Code Tester, and Project Manager—into four sequential phases of design, optimization, testing, and documentation. By mirroring human software and algorithm-development processes, these agents collaborate through atomic interactions to complete NLP2Coding and algorithm evolution. Experimental results reported by the authors show that algorithms generated by AgentAD outperform state-of-the-art human-designed counterparts across EOSSP scenarios. This work is particularly important because it illustrates a shift from LLM-assisted engineering toward LLM-orchestrated engineering design, in which models do not merely describe solutions but actively participate in creating and improving them.

Several articles further demonstrate how domain-specific LLMs and LLM-powered agents can support operational decision-making. We also want to point to a previously published article by Fan et al., which can be read in the Volume 60, 2026 of Engineering. The authors propose LLM4M, a domain-specific LLM for wind farm maintenance decision-making through labeled-data-supervised fine-tuning. By learning from mathematical programs for maintenance planning, LLM4M generates maintenance strategies under various failure modes and cost settings, with a reported error approximately 2% from the optimal strategy and maintenance cost deviations of approximately 5% when strategies are correctly generated. Sheng et al. extend LLM applications to online education by constructing heterogeneous concept graphs for cognitive diagnosis. Their model uses LLM-driven and retrieval-augmented relation mining to identify prerequisite, parallel, and synergistic relations among knowledge concepts, thereby improving the assessment of student mastery. Xie et al. integrate a ConvLSTM-based fire situational awareness model with an LLM-powered emergency response agent. Their framework reconstructs building temperature fields from partially failed sensor networks and generates user-specific operational recommendations for trapped occupants, firefighters, and commanders during dynamic fire emergencies.

Industrial design automation provides another important direction. He et al. present LLM-IDA, an LLM-driven framework for industrial design automation that combines conceptual design, knowledge-based design, and digital prototyping. Through a multi-tiered AI agent group and the four modular processes of specification, quantification, surrogate modeling, and prototyping, LLM-IDA connects natural language reasoning with computer-aided design (CAD) and computer-aided engineering (CAE) tool invocation, physical simulation, and iterative optimization. In doing so, the framework demonstrates how LLMs can help convert abstract requirements into manufacturable and evaluable design solutions with reduced human intervention.

As we navigate through these diverse yet interconnected articles, we are invited to envision a future in which ChatGPT becomes a grounded collaborator in engineering practice. The future of LLMs in engineering will not be defined by conversation alone but by domain-specific adaptation, multi-agent collaboration, verified tool use, real-time situational awareness, and accountable human-AI interaction. This special issue serves as a timely repository of current research and a guide toward engineering systems in which language models, scientific knowledge, computational tools, and human expertise work together to solve complex, high-stakes problems.