UX Collective: becoming an AI-native designer

Published in April 2026 on UX Collective, this article by Sen Lin is a practitioner account of a workflow rebuilt around AI-native tools. Lin’s current primary stack is Claude Code, Figma Make, and LLMs, replacing what was previously a Figma-centered, static-deliverable workflow. The article covers what changed, why it is harder than it sounds, and what working this way has taught him about design skill.

The tacit knowledge problem

Lin argues that the hardest part of working with AI tools is not learning the tools themselves — it is transmitting design knowledge to the AI. AI systems understand the world’s explicit knowledge: documented patterns, named components, described conventions. What designers bring to their work is largely tacit: accumulated through years of practice, expressed through decisions that are hard to articulate as rules.

The quality of AI output depends on whether the designer can make that tacit knowledge explicit enough for the model to act on. This means better prompts require a designer to be more deliberate about what they know and why. In practice, Lin finds that vague prompts reproduce generic outputs, and that the designers who get better results from AI tools are those who have already developed strong opinions about what they are building.

The demo as a design artifact

One of the article’s core observations is that a working demo in a real browser carries a kind of persuasive weight that a static mockup does not. Lin cites a principle attributed to designer Cleo, a former Facebook designer, who calls it the “Aura of Inevitability”: when a design concept exists as running code, it takes on a sense of gravity that changes how stakeholders and collaborators respond to it. A prototype that scrolls, loads, and responds to interaction is experienced differently than a Figma frame, even when the content is identical.

This has consequences for how designers move through the feedback cycle. Lin describes getting faster, more concrete feedback from working demos than from static presentations, because the demo removes the interpretive gap between what a designer shows and what a stakeholder imagines.

What accumulates with practice

Lin notes that after building several products with AI tools, designers start to develop what he describes as a faint “sense of the material” — the equivalent of the muscle memory that comes from years of traditional craft. This is not the same as experience with legacy tools: it is the specific knowledge of how AI models interpret certain kinds of prompts, where they produce noise versus signal, and which design decisions require human review rather than AI generation.

Who should read this

Product and UX designers who have access to AI coding tools but are finding the quality of outputs inconsistent. The article is also useful for design leads thinking about how AI changes what experience means — tacit knowledge has always been the differentiator, and AI does not make it less important.