The future of generative interfaces

I saw my first processing.js demo at UCLA in 2012. Casey Reas made me question everything I thought I knew about making art. Who's the real artist – the person who writes the code, or the code itself?

Now I ask a different question: What if the interface itself could create with us?

Most AI tools today work like a waiter – you order something, wait, and get what you asked for. But imagine if your 3D modeling software watched you work and spawned new tools on the fly. Request more rooms in your design, and suddenly there's a slider to control them. The software isn't just following orders – it's building you a custom cockpit.

This isn't science fiction. Figma already previewed something like this at Config. Design a layout, and the interface grows new controls specifically for that design. Not preset options – tools born from your work.

The real magic is speed. Today's AI often feels like talking to someone on a laggy video call. But these new interfaces could feel like a video game controller – instant, natural, alive. Picture adjusting complex 3D models as smoothly as moving Mario through a level.

Projects like ComfyUI and GLIGEN are already pushing this direction. Krea lets you use anything on your screen as input, updating in real-time. We're building tools that don't just understand what you want – they evolve to help you create it.

But this brings hard questions:

• How do we keep interfaces learnable when everyone's are different?
• How do we protect privacy when interfaces learn from behavior?
• How do we make tools that empower rather than replace?

The next revolution isn't just about smarter AI. It's about interfaces that grow with us, that feel less like ordering from a menu and more like jamming with a band.

Want to build this future? Start here:

• Push computation to the edge for speed
• Think beyond text prompts
• Mix AI with old-school parametric design
• Make it feel alive

The interface of tomorrow moves and breathes. What are you going to build?