How I Program with LLMs (blog.wesleyabbey.io)

A software engineer shared a pragmatic approach to programming with Large Language Models (LLMs) in a recent report, underscoring a sustainable workflow that incorporates tools like ChatGPT, Claude Code, and VS Code. The author highlighted a classic problem-solving loop as the foundation of their process, detailing how LLMs enhance various phases—from defining problems and generating plans to task execution and self-review. Utilizing features such as dictation for speedier coding and structured JSON documents for task management, they streamline their workflow, ensuring efficiency while maintaining a robust review process. This report is significant for the AI/ML community as it illustrates practical applications of LLMs in software development, moving beyond theoretical discussions to actionable methods. Key technical insights include the use of machine-generated self-reviews, which can detect errors before human oversight, and the efficient splitting of work into manageable pull requests to facilitate easier code reviews. The author's experience reflects a growing trend in leveraging LLMs to augment human capabilities in coding, pushing the boundaries of traditional development practices while advocating for continuous human engagement in the process.