AI coding assistants shift software practice from writing code to crafting prompts. We argue that prompts should be treated as lightweight, evolving requirement artifacts that intentionally mix in information from the solution space. We propose a simple conceptual model, the Prompt Triangle, with three components: Functionality and Quality (the requirement), General Solutions (solution strategy and technology choices) and Specific Solutions (implementation-level constraints). We hypothesize that prompts evolve across these components, are shaped by user factors, stabilize on functionality over time, and that completeness across all three components correlates with code quality. We outline an empirical agenda to test these hypotheses through corpus creation, longitudinal studies, controlled experiments and the development of requirements-aware prompting guidelines.
