JHDD UI UX Report — 2026.10.04

The author of “The feed I produce myself” describes multiple AI agents working in other tabs, highlighting a growing disconnect from holistic work.

This fragmented mode of operation, where system growth occurs piecemeal and without a singular overview, characterizes a broader pattern across AI-driven product development. The perceived ease of building agentic systems leads to complex, fragile interactions that users and even developers struggle to comprehend. This rapid, unexamined accretion of AI functionalities fundamentally challenges established principles of usability, user understanding, and ethical design, creating what has been termed a “New Big Ball of Mud” where systems become unmanageable and opaque.

The mainstream industry discourse often frames AI integration as a technical scaling problem, focusing on inference speed or model costs. This view is misguided. As Walter Doherty’s 400-millisecond threshold reminds, productivity hinges on interaction pace. When product teams explain slow AI features with “the model takes eight seconds” or “inference is expensive,” they often sidestep the critical design work. The root cause of “slow AI” is frequently a failure in front-end design and a lack of defined “done” states, not inherent model limitations. It is commonly assumed that AI makes building “nearly free,” fostering a culture of unconstrained addition. This assumption is precisely what creates fragile “big balls of mud” rather than robust, usable user experiences, ultimately hindering true productivity and creating poor interaction patterns.

Erika Hall, cofounder of Mule Design, asks a critical question in “In/tension modeling”: “What gives us the right to make the interventions and choices we are contemplating on behalf of others?” This inquiry is not a niche ethical concern; it is central to designing effective and responsible AI interaction patterns. The prevailing industry push prioritizes output and novelty over deep understanding and accountability. This focus on maximizing agent utilization, as seen in the “irresponsible not to use” mentality described by the author of “The feed I produce myself,” often sidelines the crucial work of Empathy Mapping and In/tension modeling. Without visualizing user attitudes and behaviors or navigating opposing voices, designers risk creating agentic systems that cause unintentional harm. This results in products that may perform tasks but fail to respect user context or autonomy, eroding trust and fostering frustration. By mid-2028, organizations that neglect ethical inquiry in their AI design processes will face significant user trust erosion and increased regulatory scrutiny, leading to substantial redesign costs that far outweigh initial development savings.

The primary resistance to this more considered approach comes from the immense pressure for rapid deployment and continuous AI-driven production. Product teams, driven by the perceived competitive advantage of AI features and the low apparent cost of initial implementation, prioritize quantity and speed. The belief that “everyone’s model is slow” becomes an excuse to defer essential front-end work and ethical review. This creates a cycle where the very “agents working in other tabs” that drive production also prevent a holistic view of user impact and system integrity.

A working UI UX professional should this week initiate a “right-to-intervene” audit for all current or proposed AI features, directly referencing Erika Hall’s question. This involves a facilitated team session using In/tension modeling principles to explicitly identify potential harms, biases, and unintended consequences before any code is written or new interaction patterns are solidified. Simultaneously, insist on defining interaction patterns for AI features to meet Doherty’s 400-millisecond threshold, pushing for front-end optimization as a core requirement rather than relying on expensive, incremental model improvements. Documenting these ethical and performance considerations within design systems for future AI component development will ensure consistency and accountability.

---

Curated References