2026 Volume 7 Issue 3 Pages 158-171
The rapid evolution of large language models (LLMs) is dramatically improving the productivity of soft-ware development. This paper reports the practical effectiveness of LLM-assisted coding through several system development cases aimed at promoting digital transformation (DX) in the infrastructure field. Specif-ically, we present three cases: (1) development of an embedded device using an ESP32 microcontroller and a camera, (2) development of a Go backend and Web UI for a school-bus safety-confirmation system, and (3) development of an Android application for the same system. In all of these domains, a substantial reduction in development effort and the achievement of practical quality were confirmed; in Case 1, given pre-defined requirements, development that conventionally required about three months was completed in about 90 minutes, and a controlled, logged reproduction experiment reached an equivalent implementation in about 68 minutes. These findings suggest that LLM-assisted coding can serve as a ready-to-use asset for a wide range of system development in the civil engineering and infrastructure fields.