YouTube reaches billions, but creation has always belonged to the few. AI Playground changes that.

I originated this concept in 2023 and spent two years driving it through organizational resistance before it became YouTube's top design priority for 2025. Today it's shipped to millions of Shorts users, the first AI-native surface YouTube had ever built.

About the Team

I led product design on YouTube's Media Generator team alongside researchers, content designers, PMs, AI researchers, and engineers, partnering with Google DeepMind to bring cutting-edge generative models into consumer experiences across all YouTube surfaces.

My Role

I led the UX end-to-end: workshopping across YouTube leadership, iterating with PM and eng partners, prototyping with AI researchers and engineers, and ultimately building YouTube's first AI-native creation surface.

Process

This was a 0 to 1 project in every sense, new surface, new technology, new user behavior, systems built from scratch, and high visibility with leadership. The process had to stay flexible: close collaboration across 5+ teams, constant workshops to align in an ambiguous space, and design systems built to evolve as fast as the technology did

Problem

While exploring how GenAI could unlock creation for more users, our research revealed a striking gap: ~95% of users abandon the creation journey at the camera, yet 47% of Gen Z identify as creators. The blocker wasn't ability, it was vulnerability. Effort, resources, cultural expectations, and for women in particular, the very real risks of overexposure and harassment.

That's when the question shifted for me: what if GenAI could make creation feel safe? Not just easier , but a space where anyone could express their vision without fear of judgment, overexposure, or being pushed out.

User Need

Through research, we identified our core user: Gen Z lightweight creators coming to YouTube Shorts looking for easy, safe, fun ways to express themselves. Our existing product was built for the intentful creator, someone who already had an idea and felt ready to express it. It wasn't built for everyone else. GenAI offered a way to change that: through automation, guidance, and personalization, it could remove the barriers to creation and give every user what they needed to actually make something they’re proud of.

Solution

The hypothesis: a unified creative playground where every user has access to everything they need to make anything they can imagine, easy, expressive, and fun. The design challenge was making that surface feel inspiring and intuitive while staying adaptable to real constraints: compressed timelines, evolving GenAI capabilities, technical limitations, and coordination across YouTube's platform at scale.

Design Strategy & IA

Modularity was non-negotiable. We were designing for a technology that was evolving faster than any roadmap could capture, new tools, conversational UI, community features, and capabilities we didn’t yet have. The system had to be agile enough to absorb all of it without breaking, and scalable enough to grow without a redesign. The framed the structure around five key moments in the user journey: onboard, discover, navigate, learn, and connect, so that as new features are added over time, the experience stays coherent and grounded in user need rather than product complexity.

Options

We explored multiple visualization approaches for these modules and iterated extensively to identify the design format that both enhanced the user experience and inspired creativity.

Information Architecture

The framework was built as a series of interchangeable modules, each tailored to a key stage of the user experience: from setting expectations, to driving discovery, simplifying navigation, and empowering creation, and learning.

1. OOBE

   Sets expectations so users arrive knowing what Playground is

2. Hero Carousel

The primary inspiration surface and entry point into creation

Creation Flow

The hero carousel provides streamlined entry points into the creation flow for media generation.

3. Tools

A clear, scannable view of everything available

4. Examples

Curated outputs that make the blank canvas feel less daunting

Content

I partnered with the PM team to generate final product content using Veo, Google DeepMind's video generation model, providing art direction to ensure outputs reflected real user needs, not just what looked impressive in a demo.

Future Milestones

M1 delivered the MVP, core creation tools (image, video, music, effects) and a modular system built to grow. The roadmap ahead includes open prompting, animation, remixing, and editing, all designed into the architecture from day one.

AI Playground launched as YouTube's first unified AI-native creation surface, bringing video, image, motion, music, and effects together for the first time, reaching 115M+ creators and becoming YouTube's top design priority for 2025.

Reflection

Designing for AI creation at this scale meant constantly navigating the tension between possibility and accessibility. The most important design decisions weren't about what the technology could do, they were about what to show, what to hide, and how to make a blank canvas feel like an invitation rather than a barrier.