Bringing Stories to Life with Props, at Scale

Over the summer I joined the Reading & Writing team to work on an unreleased, more visual version of Duolingo Stories.

My task was to explore ways to add “props” to these stories, bridging the gap between what learners read and what they see on screen. The catch was that every scene needed to be assembled automatically, so bespoke designs were out of the question. Whatever I designed had to work across thousands of scenes without any human oversight.

So I built a rules-based system. Collecting data on how often props appeared and in what forms, I sorted every prop into a small set of categories, designed specialized treatments for the most common ones, and worked through the edge cases that could break it. I documented the full system and presented it in a cross-functional handoff with engineering, PM, design studio, and design managers, giving the team a strong foundation to build off during implementation.

Early on, my PM called this project “an edge case goldmine,” and he wasn’t kidding (ᵕˆヮˆ).

Just when I thought I had the framework figured out, another script would come along and break it. I kept coming back to one of the first questions I asked, which seemed deceptively simple: what even counts as a prop?

I expected a project this rules-heavy to feel restrictive, but it ended up being one of the most creative things I worked on all summer! My favorite part was determining which rules to hold on to and which ones to stretch.