Ryo Lu, Head of Design at Cursor, explains that unlike the past when software development was fragmented by roles, the era has arrived where designers can directly work with code and immediately implement their ideas through AI tools like Cursor. He emphasizes that since AI quickly produces generic output, human "taste" and "choices" become even more important, and that software should be not just a tool but a "toolbox" that allows users to configure their own environment. Drawing on system-centric philosophies like Notion's, he discusses how AI interfaces maintain simplicity while inheriting core computing concepts dating back to 1984.
1. The Dawn of an Era Where Anyone Can Design and Develop
Looking back over the past 15 years or so, the art of making software became highly fragmented. Roles were divided — PMs, designers, developers — each using different tools and languages. But the emergence of AI tools like Cursor is reversing this trend, making the concept and skills of design accessible to far more people.
In the past, turning an idea into an actual product required understanding all the complex layers of abstraction, or assembling a team of specialists across various fields. A designer would create mockups in Figma, a PM would write documents, meetings with developers would follow, and the actual product might take months or even a year to materialize — often deviating from the original intent. Now, even with just a vague idea, you can hand it to an AI agent and immediately see a 60-70% complete result.
The feedback loop has collapsed. We've shifted from "the stage where you need to understand every concept to build software" to "the stage where you build something first and iterate quickly, even if it's imperfect." (...) What matters most is that AI knows the "codebase" — the material and truth from which we build software.
Now designers and non-developer roles can implement their ideas directly in code with AI assistance. This goes beyond just making developers code faster — it means the opportunity has opened for anyone to become a "Builder."
2. Role Boundaries Collapse, Unifying Around the Truth of "Code"
In traditional software development, each role was siloed. Designers were locked in Figma files, PMs in Google Docs, data analysts in yet another tool. Tools like Notion tried to unify them, but changing people's inertia wasn't easy. AI and Cursor are attempting to solve this problem by connecting and absorbing all artifacts.
Building software is ultimately about modifying code. A PM's spec or a designer's mockup is merely an indirect means of changing code. Cursor reduces unnecessary translation layers and miscommunication by sharing the "Truth" of code at the center of all these processes.
For an engineer, it might be a code editor, but for a designer, it could be a visual tool, and for a PM, it might look like a document tool. But they're not using disconnected apps. (...) You just think of an idea and iterate in whatever way is most comfortable for you.
This eliminates the need to ask "Where's the design file?" because the AI agent already understands the codebase and project context, dramatically increasing collaboration speed and efficiency.
3. Why "Taste" Becomes More Important in the AI Era
In a world where AI can both code and design for you, what's left for humans? Ryo defines it as "Taste" — or "choice." AI (LLMs) have been trained on all the world's data, so they can produce technically impressive results in an instant. But since AI cannot form its own opinions, without human intervention it only produces average output.
If you don't input your opinions, AI will just produce "AI slop." (...) Because AI has seen everything, it can perform the baseline really quickly and well. What needs to go on top is your taste — your choices about what's good and what's right, drawing your own boundaries.
Ultimately, the core competency of future creators will be the ability to select what's "good" from the countless options AI generates, based on their own standards and experience.
4. Design Is Not Aesthetics — It's Systems Design
Ryo shares his deep design philosophy drawn from his experience at Notion and Asana. He doesn't view design as merely making buttons pretty or adjusting pixels. For him, design means designing the simplest and most powerful concepts.
For example, Notion may look like a notes app on the surface, but it's actually a system built on a few core concepts: "blocks," "pages," and "databases." When these concepts are well-designed, users can combine them to build websites, project management tools, and more. In contrast, "selfish apps" built for specific purposes only grow more complex as features are added, eventually losing their initial simplicity.
Design isn't just about deciding on 6-pixel or 4-pixel border radius. Design is about designing the simplest system that does the most for the most people, with the fewest concepts and the fewest code paths.
Cursor follows this philosophy as well. While it may appear to be a complex development tool on the surface, it contains universal, simple concepts for interacting with AI agents. Through this, users can create their own tools.
5. The Constraint of Simplicity and the Interface as a Toolbox
As AI interfaces evolve, should all software simply become a "chat box"? Ryo says no. An empty input field sitting there alone can feel overwhelming to users. Instead, AI should exist as a universal interface that manifests in various forms suited to the user's context.
The biggest constraint is "simplicity." (...) There's a limit to the number of concepts or amount of information you can expose to users at any given moment. (...) Designers are no longer the people who decide "where the button goes" — they need to think about the minimal set of abstractions and systems that can handle all these variations.
Cursor is like a "toolbox" that lets users customize the tool to match their workflow. Some prefer viewing code directly; others prefer visual previews. What matters is that the system flexibly accommodates this while keeping the defaults simple.
6. We've Been Building the Same Thing Since 1984 (RyoOS)
Finally, Ryo introduces his personal project RyoOS and discusses his sources of inspiration. Rather than just creating design mockups in Figma, he writes code and prototypes directly to find inspiration. RyoOS is a project that recreates the feel of retro operating systems like Macintosh System 7 or Windows 95 on the web.
We've been doing essentially the same thing since 1984. (...) The only difference was the technical constraints. Drawing pictures in Paint, writing text in a text editor, having icons and a desktop. Almost nothing has changed.
He collects old Macs and iPods, discovering timeless value in past interfaces. Technology has advanced, but the fundamental concepts and patterns of how humans interact with computers haven't changed much. Even future-oriented design is ultimately just the process of reinterpreting familiar past concepts for a new medium.
Closing Thoughts
Ryo Lu says AI is breaking down the boundary between designers and developers, returning us to an era of "Complete Builders." The technical barriers have lowered. What matters is not which tool you use, but what you want to build with it, and what "taste" you want to infuse into it — a fundamentally human question. What creators of 2025 need may not be complex technical knowledge, but the insight to see through to the essence of things and a perspective that is uniquely their own.