Product design · 0 → 1

Yedoo

An AI reading app that helps people understand a book in depth. Readers ask, discuss, and create with AI, and pick up the same book on iPhone, iPad, Mac, or the web.

Role
Product strategy · UX/UI · Design system
Scope
App preview · Landing page · Logo
Platforms
iOS · iPadOS · macOS · Web
Built with
Claude Code · Codex

My contribution

I led the product from strategy and interaction design through implementation, brand, and launch assets. I defined the design system, wrote the copy, and reviewed and refined the experience.

Problem

Help readers understand a book, not just finish it.

Yedoo is my own product, an AI reading app I designed and built. It began with a design question: how could a reading app help people work through a difficult idea while keeping the book in focus?

Target reader: a heavy reader who owns print and ebooks, what they want and where they get stuck
Diagram: a reader switches between the book, notes, AI chat, web search and highlights, versus one Yedoo workspace

Constraints

  • Engineering. Too few engineers to build everything we wanted.
  • Time. Not enough time, so the scope had to stay tight.
  • Testing. No time for A/B testing, so decisions came from the app audit, design reviews, and pilot users.

The questionHow might we help readers work through a difficult idea while keeping the book in focus?

Design stance

Go deeper with AI. Readers talk ideas through with AI, turn them into images and videos, and in Round Table, invite AI versions of thinkers like Confucius, Benjamin Franklin, or Steve Jobs, each set up with their own worldview, to discuss the book together.

Read anywhere. The same book, highlights, and conversations carry across iPhone, iPad, Mac, and the web, so reading never starts over.

Inspiration

See how other apps handle the same moments.

Before designing, I collected flows from reading apps and from apps that do onboarding, upgrades, and purchase confirmation well, to learn the patterns first.

App audit

Audit the first build, screen by screen.

I walked through every flow of the early build, then called Lumo, and noted what a heavy reader would trip over.

01

Sign up

Only Google sign-in, no questions about reading habits, and no clear Free versus Pro. Nothing guided readers to upload their first book.

02

Library

Target readers already own books, so upload should come first. There was no way to organize, sort, or archive, and the tab icons and colors needed work.

03

Discover

Discover is where readers browse books outside their own library, grouped into sections. There was no search and no language or category filter, and the sections had no labels saying why a book was there.

04

Round Table

Round Table lets readers pick a topic from a book and watch several AI voices discuss it. The feature was never explained, readers could not see who the participants were, and the discussion ended with no way to join in or keep it going.

05

Empty states and settings

Empty Chat and Visuals did not say what to do next. Settings needed a close button, a way to report a problem, and contact support.

Overview of every screen in the first build, grouped by area: sign in, library, discover, book detail, chat, Round Table, visuals, reading, and settings, with notes beside each
Every screen of the first build, grouped by area, with notes beside each.

Process

Turn a broad concept into clear design rules.

AI help needed to be easy to reach without interrupting reading. I kept the book at the center, placed questions beside the passage, and gave discussion and creation their own entry points.

01

Give reading and AI distinct roles

The library and reader prioritize returning to the book. Chat, Minds, and Creations remain available as a separate layer, balancing access to AI with a focused reading experience.

02

Keep AI beside its source

I made questions contextual to a passage, so readers can move between the book and an explanation without losing their place.

03

Define the system before scaling

I connected type, components, and interaction rules through a shared design system, then adapted the layout to each platform's space and reading context.

Early AI-generated concept for the Creations web experience
Starting conceptClaude Design supported exploration; I defined the navigation, interactions, copy, and system. Claude Code and Codex supported implementation.
Yedoo design system type tokens in Figma
Design systemReusable type and component rules connected through Figma MCP.

Product Design

Read, ask, discuss, and create, on every device.

Readers can ask about a passage, explore other perspectives, or create something from the book without starting over elsewhere.

Yedoo macOS home with reading progress and recommendations
macOS
Yedoo library on iPad and iPhone
iPadOS and iOS

iOS

Prioritize returning to a book, reading, and listening on the go.

iPadOS

Use the extra space for browsing the library and creations.

macOS

Keep reading and understanding tools visible in one workspace.

Web

Support web reading and shareable highlights.

The reader keeps the text first. Marking and navigation tools appear only when needed.

As shipped
iOS Create page with type picker13
Creations sheet with templates12
Prompt and generate settings35
Recent creations with a generated image4
Full screen image viewer4
  1. 1
    The type is picked twice

    Image is chosen on the Create page, then the sheet asks again.

  2. 2
    Too many decisions in one sheet

    Type, template, Skill, prompt, and size share one long sheet. Generate sits below the fold.

  3. 3
    Names don’t match

    Mini-book becomes Mini, credits are called insights, and the empty state asks for a book that is already chosen.

  4. 4
    Controls cover the result

    Buttons sit on top of the image, and the full screen view offers only Close.

  5. 5
    Waiting is unclear

    It takes one to two minutes, but nothing says whether readers can leave or will be notified.

  1. Redesigned Create screen with book, type, template, prompt and Generate on one screen
    01Pick and write

    Book, type, template and prompt sit on one screen. The type is picked once, and size sits next to Generate.

    Fixes 1 and 2
  2. Creations page with a generating card above the earlier version
    02Keep reading

    Generation runs in the background. A card in Creations shows the progress and tells readers they can leave. They get a notification when it is ready.

    Fixes 5
  3. Full screen result with versions and Save, Share, Edit prompt and New version below the image
    03Look, then act

    The image gets the screen. Versions and actions sit below it, and the name stays Visual brief from template to result.

    Fixes 3 and 4
  4. Versions sheet listing twelve versions with kept versions at the top
    04When there are many versions

    Tap the version label to see every version in one sheet. Each row says what changed, and kept versions stay at the top.

    Scales step 3
  1. A scanned PDF has no text

    The AI has nothing to read. Import flags it, so the book still opens, and the AI tools say they are limited.

  2. The book and the question use different languages

    A reader asks in Chinese about an English book. The answer follows the reader, and quotes stay in the book’s language.

  3. A long generation fails

    The reader may have left. The card says what failed, offers a retry, and the attempt does not use a credit.

Results

From concept to an iOS launch in one month.

I designed and built Yedoo, then launched the iOS app in one month. It is now live on iOS, iPadOS, macOS, and the web. In the pilot, 80% of users completed sign-up, and paid users were about twice as active as free users.

Pilot sign-up
80%

Of pilot users completed sign-up.

Paid user activity
~2x

About twice as active as free users.

Launch
Live on iOS
App Store rating
5.0

Early ratings since launch.

Yedoo App Store preview showing the product identity, library, book detail, and reader
App Store visuals carried the same product identity from the app into its launch.