The rapid integration of artificial intelligence into consumer and enterprise software has created a growing gap between the sophistication of AI capabilities and the interfaces through which users interact with them. While AI models have advanced considerably, the dominant interface paradigm - dashboards, forms, multi-step wizards, and navigation hierarchies - remains largely unchanged from the pre-AI era. The most common adaptation has been the addition of conversational chat interfaces, which this paper argues is insufficient. This paper presents The Interface Manifesto, a set of ten design principles for AI-native user interfaces derived from practitioner experience building AI-powered assessment tools. The principles address fundamental shifts required in interface philosophy: from navigation to intent inference, from data density to synthesized clarity, from user configuration to opinionated defaults, and from engagement optimization to trust-based design. Each principle is presented with its theoretical grounding, practical implications, and an evaluative test for product teams. The work extends foundational contributions by Amershi et al. (2019) on human-AI interaction guidelines, Nielsen's intent-based outcome specification paradigm (2023), and Weisz et al.'s generative AI design principles (CHI 2024). Supplementary material includes before/after visual comparisons for four of the ten principles.
Gunjan Thapa (Thu,) studied this question.