JHDD UI UX Report — 2026.10.12

Larry Tesler, known for his ‘NO MODES’ license plate, spent his career eliminating confusing states from software interfaces.

The recent discussions surrounding the new Siri, Claude’s context libraries, and Tesler’s Law reveal a quiet but significant shift in interface design. Software complexity, once managed through explicit modes or visible controls, is now increasingly shunted into an opaque layer of AI context. Users are expected to implicitly manage, or even create, intricate contextual information for AI agents, often without adequate tools or clear understanding of the AI’s limitations, as highlighted by observations that Claude power users struggle to find or edit their critical context files. The interface itself has become simpler, but the mental model required to effectively operate AI systems has grown more convoluted, transferring the burden of understanding from the system to the individual.

The case of Siri exemplifies this shift. Its biggest strength, according to analysis, is not raw intelligence but its ability to leverage relevant personal context. However, the user experience does not sufficiently equip individuals to understand or manipulate this context effectively. The prevailing industry narrative suggests that AI simplifies user interactions by abstracting away underlying complexity. This perspective misses a critical detail: AI often merely relocates complexity, making it less visible rather than truly removing it. Unlike the clear visual modes Larry Tesler worked to eliminate at Xerox PARC with Gypsy, or later at Apple, the “modes” of an AI system are now embedded in its ever-evolving context library, an interaction pattern that remains largely unaddressed by design systems. Designers often create a superficial “no modes” experience, without fully grasping that new, invisible modes have been introduced through context engineering. This leads to friction, where users either underutilize powerful AI features or become frustrated by their unpredictable behavior.

The mainstream belief that AI inherently reduces cognitive load for all users by handling tasks autonomously is a misdirection. While surface interactions may appear simpler, the underlying demands on user mental models for managing context, understanding AI limitations, and predicting system behavior have increased significantly. By mid-2027, this lack of transparency in AI’s contextual operations will necessitate a new standard for explainable AI interfaces. Design systems will begin incorporating explicit patterns for context visualization and editing, not for power users alone but for everyday interactions, treating context management as a first-class citizen in the UI, akin to how file systems are managed today.

The resistance to this transparency often comes from product teams prioritizing perceived simplicity and “magic” over genuine user control and understanding. Marketing narratives frequently frame AI as a seamless, effortless intelligence, which disincentivizes exposing the intricate mechanisms of context engineering. Furthermore, the technical challenges of exposing and allowing manipulation of complex AI models are often deemed too high, leading to a default position of hiding these functionalities from the user.

UI UX professionals should conduct user research specifically focused on how users conceptualize, manage, and debug AI context in their daily workflows. This means moving beyond traditional task-based research to explore user mental models of AI’s “memory” and “understanding,” identifying pain points in modifying or overriding AI’s inferred context. Designers should prototype and test interaction patterns that enable users to view, edit, or reset relevant contextual data, even for seemingly simple AI assistants, rather than defaulting to black-box implementations.

---

Curated References