Designed & Created by Jamie Chu with Claude AI
Prototype scope. Two paths in one course. In the first, a pair starts Japanese from nothing. In the second, a pair already learning Japanese is a unit apart and decides to sync up. One Korean screen exists to establish a second course. Every screen is a capture from a real account, so the profile and the buddy are real people. Simulate controls stand in for time that would pass in reality: completing a unit, a buddy answering, reaching a milestone. Static fidelity, dark mode throughout. Not built: notifications, feed posts, the profile roster and menu, ending a pair, and anything outside those two paths.
Contents
Feature Strategy & System Design
Two friends going to Japan in six months. A couple where one is learning the other's first language. Two people who both quit and want to try again together. Each of them comes to Duolingo with a specific person in mind, and the app can give that person a number and nothing else.
That is a consequence of how the social layer was built, not an oversight. Friend Streak counts the days two people both practised. Friends Quest pools two people's XP toward a weekly goal. Leaderboards rank XP across a cohort. All three work whether you're learning Japanese and your friend is learning French, and that indifference to course is what lets each of them reach the whole user base. The price is that none of them can put two people in the same material. What a pair shares is a count.
Course Buddies gives up that reach on purpose. It works only when two people are in the same course, and in exchange it lets them work through the same units from the same starting point, quiz each other on what they've both covered, and reach milestones together. The person you're learning with knows exactly what you're working through.
Casual learners are the target. They have the habit without the consistency: roughly eight active days a month, with the app installed and notifications on. They can be reached, which the other groups can't say, and the distance from eight days to fifteen is the largest lift available anywhere in the population. A quiz waiting from a specific person is what closes that gap on the days a solo streak wouldn't.
Lapsed learners are the more vivid case and the harder one. The per-user lift is bigger, since they're starting from zero, but nothing in the app can reach them. The only route back is a friend asking, which the invite flow is built to serve.
First-time accounts are the cleanest fit for the mechanic. No prior progress, no rewind, a shared Unit 1. Every day they show up is additive. A new account follows nobody, so pairing depends entirely on a conversation that happened somewhere else.
These are three groups, not four. Someone casual who picks up a second language is still a casual learner, already counted, with no separate story.