Re-Architecture of Proprietary Software Suites into Open-Source Alternatives

Abstract

This paper examines the paradigm shift in software engineering from authoring code line-by-line to directing AI systems that analyze, specify, and translate existing system behavior. Specifically, it analyzes how large language models (LLMs) such as Claude are utilized to study the observable behavior of proprietary software suites—such as Adobe Creative Cloud and Microsoft Office—and reconstruct their functionality as modular open-source alternatives. The study covers three core technical threads: The application of long-context LLMs in deconstructing UI/UX patterns, rendering pipelines, and file formats (PSD, DOCX, etc.). The integration of the Model Context Protocol (MCP) to establish real-time, closed-loop interaction between LLMs and live open-source codebases, test suites, and execution environments. The engineering methodology of interface-level clean-room reverse engineering. Additionally, the paper analyzes GitHub as the coordination infrastructure and automation layer for distributed, AI-assisted development, concluding with a case study of the open-source "ArtCraft" project ecosystem and a discussion of legal, technical, and governance challenges.