The emergence of Large Language Model (LLM) coding agents has created a new development paradigm where a single developer directs autonomous agents to design and implement software systems at unprecedented scale and velocity. Despite widespread adoption, no formal methodology exists for this practice, and empirical data on its productivity characteristics remains scarce. This paper introduces Agent-Driven Design & Development (ADD), a structured methodology that formalizes the full pipeline from idea to production using LLM agents - covering both the design phase (specifications, architecture, patterns, documentation) and the development phase (code generation, testing, integration). We present empirical evidence from 8 real-world projects spanning 3 programming languages, totaling 186,546 lines of code produced by a single developer over 30 active development days.
Filipe Mendes (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: