Dive Club
Posts
🤿 designing the future of Cursor

🤿 designing the future of Cursor

3 big ideas from Ryo Lu

Ridd
July 11, 2025

Cursor is easily one of the most impactful companies in all of tech right now.

So this week we're going behind-the-scenes with their Head of Design, Ryo Lu (who you might remember from this interview when he was the founding designer at Notion).

It's a deep dive into systems thinking, agentic design, and Ryo's journey building his ideas with AI.

Some highlights:

What Ryo believes about the future of HCI
How Ryo handles a barrage of user feedback
How Ryo approaches systems thinking at Cursor
Laying the foundation for the future of Cursor’s design
Inside Ryo’s design process and how he knows what to pursue
Why Ryo thinks specialization will matter less in an AI-enabled future
+ a lot more

Listen on YouTube, Spotify, Apple, or wherever you get your podcasts 👇

🤝 WITH PAPER

There are a lot of people saying that you gotta become a design engineer…

But I don’t think that’s true. Because Paper is giving you a lot of the benefits of code without sacrificing the design experience

You can still make a mess and explore like crazy in an interface that feels like a streamlined version of Figma.

But the canvas itself is rendered with code not vector.

That means developers can ship your designs exactly as you designed them. No more lost in translation….

And soon you’ll even be able to import code components directly onto the design canvas so you don’t have to deal with duplicate sources of truth anymore!

It’s a pretty big deal and just another reason why I’m all-in on Paper as the next great design tool

Click the link to try it out today 👇

🎓 KEY TAKEAWAYS

My takeaways from Ryo

1 — Sculpting > Planning

Previous eras of software felt like you had to lock in the “shape” of a feature before touching the material (PRDs, task lists, wireframes, etc.).

But Ryo says AI has flipped the process upside down:

“[Software creation] is almost like sculpting. You get something then you poke at it. Maybe you get rid of certain parts. Maybe you rearrange certain parts. Maybe you say, ah, do it again, again, and again. Do it five times. Pick the right one. It changes completely… you’ve flipped how you built.”

Ryo Lu

I think that’s my main takeaway from Wednesday’s video where Ryo gives a behind-the-scenes of his vibe coding prompts. His first prompt had nothing to do with the features he wanted to build. It was simply defining the block of clay he wanted to mold.

This comes up again in next week’s episode about Perplexity—software creation is all about racing to get something in code to play with (vs. “figuring out” the design up front).

2 — Systems > Screens

My favorite thing to talk to Ryo about is systems thinking (definitely check out his Notion episode or this mini one on systems thinking ICYMI).

So I wasn’t surprised when he told me his very first project at Cursor was rethinking the core primitives and simplifying a bunch of scattered concepts into a single “agent mode”.

“Instead of five discrete little things, you just make the circle bigger... you don’t have to get rid of things to make something simple, you just have to unify them”

Ryo Lu

Ryo doesn’t think in terms of layouts or features… he thinks in primitives and how they map to the larger system.

How few building blocks can you get away with?
How simple can their mental models be?

Take Ryo’s recent work on Agents for example…

It was very tempting to treat each “chat” in cursor as a “task” for an agent to perform. On the surface this feels correct, right? No new concepts, just repurpose an existing one.

But usage patterns told a different story: some people stuff everything into one chat, while others naturally split work across many.

Ryo realized chats couldn’t reliably serve as tasks so he introduced a new primitive: the to-do list.

The best building blocks unlock future possibilities without requiring new primitives. To-do lists are the perfect example.

Maybe in the future a core use case for Cursor is non-technical people reviewing to-dos in a separate web view without ever having to enter the IDE 🤔

That’s what systems thinking looks like in practice.

3 — Malleable > Generative

A year+ ago I was fascinated by dynamically generated interfaces.

But now I’m thinking the real opportunity for designers isn’t auto-generating screens—it’s crafting the right set of primitives so users (and AI) can remix them into something that feels personal.

“The ideal interface is different for every single person… maybe in the future AI can kinda reconfigure the entire Cursor for you… but arbitrarily generating UI even the creators can’t predict is not a good thing. It just creates chaos.”

Ryo Lu

This is where malleable software comes in 👀

The designer’s job starts to feel less like defining UIs and more like shaping the blocks that AI can wield.

“you need to design the low-level architecture where you know you can fit all these different configurations and customizations. How much do you want to offer to different people? What are the different defaults people get?”

Ryo Lu

The set of blocks is fixed. But the combinations are nearly infinite which is how Ryo thinks Cursor can achieve personalization at scale.

I’ve been thinking about this a lot for Inflight lately 🤔 maybe each team can customize the core functionality in a way that feels bespoke for how they work.

What are the fewest number of primitives that we would need to enable that?

That question feels more important now than at any point ever in designing digital products.

“the things we design as designers go up one level. Instead of designing exactly how this piece of UI will look, the order of buttons, where they are placed, etc… you are actually designing a container and the patterns that make up a whole system”

Ryo Lu

Design is fun :)