Compete and Learn? Understanding Competitive Programming IDE Workflows and Designing Context-Aware AI Assistance to support Learning and Assessment

Competitive programming platforms prioritize rapid submission and automated judging, but their built-in IDEs often hinder learning-oriented workflows and lack principled AI integration. In preliminary research with competitive programmers (N=74), we found three recurring challenges: workflow fragmentation between local and platform editors, inadequate debugging and test-case visibility, and context-dependent expectations for AI assistance that differ sharply between contest and learning scenarios. Drawing on these observations and HCI principles of cognitive load reduction and scaffolded learning, we present a conceptual IDE design that uses mode-based interaction constraints to govern AI assistance across contexts. Our design separates Contest and Learning modes, introduces inline test-case visualization and error localization, and constrains AI behavior to preserve assessment integrity while advancing pedagogical goals. We contribute empirical evidence of workflow breakdowns in competitive programming environments, design goals for addressing these challenges via interaction design, and a mode-based interaction model for context-appropriate AI assistance in dual-purpose educational platforms.