Curation, Not Generation

Curation, Not Generation defines the modern designer's role in an AI-accelerated world. It means shifting from the painstaking creation of every pixel and line of code to the critical evaluation, refinement, and strategic guidance of AI-generated design outputs. Think of it this way: the AI handles the mechanical grunt work, spitting out a "confident draft" of a working interface, as seen with Claude Fable 5. Your job is no longer to just make the thing, but to make sure the thing is right. This involves a sharp eye for detail, a deep understanding of your design system, and an unwavering commitment to user experience and brand integrity.

With tools like Fable 5, the model can now take a Figma screenshot and return token-aware, system-leaning markup that faithfully respects your design. This significantly reduces the "cleanup tax" that plagued earlier code generation tools. Last year, you got a "vibes export" with invented spacing and hardcoded hex values. Today, you get a credible starting point. But a starting point is exactly that: a beginning, not an end. Curation means you are the quality gate, the arbiter of taste, and the guardian of the system. You are checking that the spacing relationships are exact, that every value maps to a token, and that the component repetition is consistent. You are ensuring the AI's output aligns with the strategic intent and the nuanced requirements that only a human designer can fully grasp. It is about elevating a functional output into an exceptional user experience, catching the subtle errors, and injecting the human touch that makes a product truly shine.

Curation, Not Generation is not passive acceptance. It is not a designer simply rubber-stamping whatever the AI spits out. This is not a job for someone who lacks fundamental design skills or a deep understanding of user needs, accessibility, or interaction patterns. It is far more than mere proofreading; it is a systemic review and a strategic intervention. The AI, even one as advanced as Fable 5, operates on the limits of its input. It cannot infer states it cannot see, like empty states, loading skeletons, or complex error conditions. It has no opinion on whether a feature is worth building in the first place, nor does it possess inherent brand judgment or taste.

Therefore, curation is not delegating strategic decisions to an algorithm. It is not an excuse to skip defining your design system; in fact, a robust system becomes even more critical as the AI leans on it for token mapping and component usage. It is not about generating from scratch without any human input; the "screenshot-to-code" workflow still begins with your meticulously crafted primary design. Nor is it a replacement for understanding the "why" behind a design. Fable 5 will give you a button, but it will not tell you if that button should be red or blue for your specific brand, or what happens when the network drops after a user clicks it. The model collapses the gap between design and code, but it does not collapse the gap between good and right. That gap is where the designer's value now resides.

Consider a common task: building a new user profile screen for an application. Pre-Fable 5, this was a multi-stage, often painful process. A designer spent days in Figma, crafting the primary state, adding redlines, and writing a detailed spec for every interaction, empty state, and error condition. This artifact went to an engineer, who spent another week translating it to code. The designer then reviewed the coded implementation, often finding pixel discrepancies, hardcoded values, or missed details, leading to costly iterations.

With "Curation, Not Generation" and Fable 5, the workflow transforms. The designer still defines intent: the screen's purpose, key user flows, and necessary states (loading, empty, error, editing). This strategic thinking, the "what to build," remains human. Next, the designer crafts the primary state in Figma, leveraging the established design system with its tokens and components. This is focused, high-leverage design work.

Then comes the AI. The designer screenshots that Figma design and feeds it to Fable 5. Within minutes, Fable 5 returns a working, token-aware React component, faithfully replicating the visual design. This is the "generation" part. Now, the "curation" begins. The designer meticulously reviews the generated output. They check for pixel-perfect fidelity against the Figma design, ensuring every spacing unit and typographic scale is respected. They verify all colors, fonts, and dimensions map to correct design tokens, catching any AI-hallucinated hardcoded values. Accessibility is paramount: is the focus order correct for keyboard navigation? Are ARIA attributes applied appropriately? Is contrast maintained? These are often AI blind spots. Since the AI only saw the primary state, the designer must ensure empty states, loading skeletons, and specific error messages are handled. While Fable 5 builds static elements, the subtle feel of an animation, tooltip timing, or validation logic still requires human oversight. Finally, the designer assesses the underlying code quality: is it clean, semantic, and maintainable for engineers?

The designer then either refines the code directly or provides precise feedback to engineering, focusing on identified gaps. This iterative refinement, driven by human judgment, ensures the final working artifact is not just functional, but truly excellent and aligned with all product requirements. The deliverable shifts from a static Figma file to a working, reviewable artifact embodying both AI efficiency and human intelligence.

Embracing "Curation, Not Generation" is not a universal panacea; it is a strategic shift best applied under specific conditions. Lean into this mindset any time an AI can reliably generate a credible first draft of a visual or interactive element. This is powerful when accelerating the translation of static designs into working code, as Fable 5 demonstrates. If you have a robust design system, AI tools can leverage it for token-aware, consistent outputs, making curation highly efficient. It is ideal for rapid iteration on visual concepts, validating ideas in working form faster than traditional methods. When the primary constraint is speed for known patterns, and you have critical judgment, curation is your superpower.

However, clear boundaries exist. Do not default to AI generation when output is consistently poor, requiring more cleanup than starting from scratch. While Fable 5 raised the bar, not all AI tools are equal. Avoid it for highly novel interactions, complex data visualizations, or new component patterns AI struggles to infer. The AI lacks "taste" or brand judgment for unique creative challenges; these demand direct human craft. Crucially, do not use this approach if you lack the critical judgment to evaluate AI output against design principles, accessibility, or system constraints. A confident but incorrect AI output is more dangerous than none. Furthermore, when building a design system itself, foundational components need human craft, not just AI generation. Finally, if your input is ambiguous or incomplete, the model is prone to "hallucinated confidence," producing plausible but wrong outputs difficult to curate effectively. Curation's power lies in the quality of both input and human oversight.

When generation is free, judgment is the product.