Show HN: A Prompting Framework for Non-Vibe-Coders (github.com)

Projex is a new framework designed to facilitate efficient prompting of large language models (LLMs) for non-developers, enabling collaborative agentic development without reliance on traditional coding skills. Unlike vibe-coding or autonomous agents, this approach promotes structured interaction with LLM models, where users can supervise and guide the development process. Projex documents serve as self-contained units of work that contain plans, evaluations, and memory, significantly reducing the memory and context retrieval demands on LLMs. This system allows for iterative development through small, manageable steps, where each projex document can be referenced in subsequent sessions. The significance of Projex lies in its emphasis on traceability, memory persistence, and structured execution workflows that enhance the decision-making capabilities of users. With built-in mechanisms like context adjustments and guardrails in templates, it minimizes errors and improves the learning curve for users interacting with LLMs. It also integrates seamlessly with Git, allowing for collaborative and organized workflows tailored to various project needs, ultimately reshaping the experience of working with AI tools in software development and beyond.