Journal of Advanced Robotics, Autonomous Systems and Human-Machine Interaction
From Intent to Code: An AI-Powered Paradigm for Translating Declarative Specifications into Optimized Implementations
Abstract
Ndenga Lumbu Barack
I present “Decrypt,” a novel intent-aware programming paradigm that directly addresses the critical misalignment between a user’s expressed goal and the code generated to achieve it. Current conversational AI for code often fails to capture the full nuance of a user’s intent, leading to frustration and inefficient feedback loops. My work introduces a direct “intent-task matching” architecture that externalizes the AI’s understanding of the programming task before code generation begins. By allowing users to inspect and intuitively refine this intermediate representation, Decrypt ensures that the final, contextually-optimized implementation is a precise reflection of the original goal, drastically reducing cognitive load and development time.