My LLM codegen workflow atm
2025-07-31
Harper Reed outlines a three-phase LLM-assisted software development workflow: (1) idea honing through conversational prompting to generate a detailed specification, (2) planning using reasoning models to break the project into properly-sized, test-driven implementation steps, and (3) execution using LLM codegen to implement each step incrementally. The approach emphasizes discrete loops, comprehensive planning before implementation, and iterative refinement to avoid common pitfalls like wasted time and orphaned code.
Was this useful?