Incrmd: Incremental AI coding by editing PROJECT.md (github.com)

INCRMD is a novel experimental framework that redefines the interaction model between developers and AI coding tools, such as Codex and Claude Code. Unlike traditional methods, where developers must input specific requests through chat prompts, INCRMD streamlines the process by allowing developers to edit a single file—PROJECT.md. The system continuously monitors changes made in this file and relays only the differences to the AI, ensuring that the model has a clear, cohesive view of the entire project specification. As PROJECT.md evolves, the clarity of project intentions increases, reducing the risk of miscommunication and contradictory prompts. This approach is significant for the AI/ML community as it addresses a common pain point in AI-assisted coding: the degradation of context over time. With INCRMD, the precision of project descriptions actually improves as more details are documented, allowing the AI to maintain a consistent understanding of the project’s goals. By relying solely on PROJECT.md, developers can focus on refining their specifications rather than navigating through multiple files and fragmented instructions. As an early experiment released under the MIT license, INCRMD invites feedback and further exploration within the AI coding landscape, promising enhanced collaboration between human developers and AI systems.