Back to workrichard.vinueza

Huma Legends

Learn English and Spanish vocabulary through illustrated stories from around the world.

role
Product, design & iOS development
period
2023 — present
platform
iOS
stack
Swift · React Native · Python · Convex · Stripe · AWS · Railway · Cloudflare

The problem

Vocabulary apps drill words out of context. People memorize lists, forget them a week later and quit.

Huma teaches words inside a story: a legend from a real place, told with illustrators from different countries and, alongside them, video and AI-made pieces. Each new word appears where it matters.

What I built

  1. 01StoryA legend from a real place, one chapter at a time
  2. 02Words in contextTap any word, hear it, keep it
  3. 03ReviewSpaced repetition on what you kept
  4. 04ProgressStreaks and a map of visited places

Key decisions

Native SwiftUI instead of cross-platform

why
Reading and audio have to feel instant. Native gave the smoothest text rendering and offline playback with the least code.
cost
Android has to wait.

Convex as the backend

why
Realtime sync of progress between devices came for free, and the schema lives next to the client code.
cost
A smaller ecosystem than Postgres; I keep the data model simple on purpose.

Stories before features

why
Every release ships at least one new legend. The content is the product; features follow what readers ask for.

Illustrators from each place, not one house style

why
A legend from Ecuador should look Ecuadorian. Local artists bring that; video and AI-made pieces fill the gaps and keep the release pace.
cost
Less visual uniformity across stories, more coordination.

Architecture

  1. 01SwiftUI clientReader, audio, review engine, offline cache
  2. 02ConvexAuth, progress, story catalogue, realtime sync
  3. 03StripeSubscriptions and receipts
  4. 04Content pipelineIllustration, video, AI pieces and audio processed at build time

What I learned

Short, illustrated chapters beat long lessons. The first version had ten-minute stories; nobody finished them. Two-minute chapters doubled completion.