Role
Product Design Lead
Credits
Leo /Product Lead, Hang /Engineer
Time
2025 - Present
+13%
DAU growth
~1M
New registered users
+23%
Daily revenue
The context
Context
Answer.AI is a learning app with nearly 7 million users, most of them middle-school to college students. It grew on one killer loop: snap a photo of a problem, get the solution, ask follow-ups. That loop brought massive traffic, but treated every student the same way. Our mission: raise retention, and grow the product into something that helps students far beyond homework.
The turning point
Why change
By 2024 the answer machine was winning, and stalling. Three forces pushed us to rebuild:
Leaky retention
Usage was transactional: come with one problem, copy the answer, leave. Exam weeks spiked; loyalty never formed.
A commoditized loop
Every competitor shipped the same snap-solve flow. Faster answers were no longer a moat.
A new AI window
LLMs had just made one-on-one teaching possible at scale. For us, or for whoever moved first.
Staying an answer machine meant staying replaceable.
The strategy
One Promise, Four Layers
The master plan
This is the blueprint I led the redesign by: one promise, delivered in four layers of product work. The rest of this case follows it in build order — each layer below is a chapter, from 01 Core to 04 Personal.
THE PROMISE
How does a shared answer machine become a school that knows each student?
Our answer to all three forces. Four layers, built from the core up.
04
Personal · Streaks plus proactive agents.
A school that remembers you and reaches out.
03
Reach · Rebuild the IP system.
Change what students think the product is.
02
Stay · Flashcards, study rooms, mock practice.
Turn a quick tool into a long-term study partner.
01
Core · Redesign Ask AI: clearer explanations, more agents, voice.
The front door. If explanations are not clearly better, nothing else matters.
Each layer stands on the one below.
Sharpen the core, build reasons to stay, reach more students, and make it personal.
Layer 01
Sharpening the Core: Ask AI, Explained Better
Why this first
Photo-solving is where every student's journey starts, so it earned the first and biggest investment. The rebuild made two bets: explanations students learn from instead of copying, and a chat calm enough to think in — every extra tool moved out of the way, into a sidebar.
Step-by-step explanations
Before
After
<>
Before — one long block. Easy to copy, hard to learn from.
After — clear steps, answer set apart.
Clean chat + sidebar
Before
After
<>
Before — a toolbar of features inside chat. Distracting, hard to stay in one thread.
After — one clean chat. A curated sidebar: key tools and history, in the student's control.
Layer 02
From Quick Tool to Long-term Study Partner
The shift
Solving one problem is a transaction; learning is a relationship. Exam prep is where the need is sharpest — the hope is that along the way, a little more actually sticks. We cut the features that didn’t earn their place, and kept the methods students already love — backed by data and social buzz.
Flashcards
The method students already know
Mock practice
Walk in ready for the exam
Online study rooms
The flexible slot — focus with company
Layer 03
A New Face: Rebuilding the IP
The perception
Students saw Answer.AI as a vending machine for answers — useful, but not somewhere to stay. The old IP didn’t help: cute but childish, and hard to remember. The new cast grows out of the Answer.AI logo and acts like a school’s favorite teachers — their personalities come through in how each AI tutor talks, so the app starts feeling like a school of their own.
Before
After
<>
Alma — the warm, outdoor-loving counselor.
Ace — the playful genius learning buddy.
Grant — the shy advisor with a head for numbers.Layer 04
Streaks, Nudges, and a School That Knows You
Make it proactive
The tutors act first. Ace, Alma and Grant get to know each student — grade, school, strong subjects, dream schools — while Ace maps mastery from follow-up questions and scanned problems. Then they reach out at the right moment: a check-in after yesterday’s tough question, a preview invite the night before tomorrow’s calculus class.
The reward loop
A light reward system keeps it going: scanning problems, mastering topics and finishing Ace’s exercises earn badges — plus random outfit drops for an avatar that shows up in the online study room.
Proactive preview
Tomorrow’s class, tonight’s heads-up from Ace.
A call with Alma
Strong in a subject? Ace books a college-planning call.
Badges to avatar
Tasks earn badges — badges dress the study-room avatar.
Honest notes
Learning From a Miss
What went wrong
Not every bet landed. In interviews, students spoke warmly about a counselor, a scholarship advisor and AP prep — and we heard a bigger need than everyday behavior ever showed. Alma, Grant and AP History quietly moved down the roadmap, and a lesson moved up: interviews are polite, behavior is honest. Now the loudest voice in every decision is what students actually do.
The wrap
Grateful, and Onward
Full loop
Design had a seat here from strategy all the way to shipping — rare, and I’m grateful to the team that made it possible ❤️ The app keeps growing on the new foundation. What I’ll remember, though, is meeting students at campus events: pulling up their report cards to show us, saying thanks, telling us it helped. That’s the best part — not the charts, but knowing they feel themselves getting better.
Results
+13%
DAU growth
~1M
New registered users
+23%
Daily revenue
Role
Product Design Lead
Credits
Leo /Product Lead, Hang /Engineer
Time
2025 - Present
+13%
DAU growth
~1M
New registered users
+23%
Daily revenue
The context
Context
Answer.AI is a learning app with nearly 7 million users, most of them middle-school to college students. It grew on one killer loop: snap a photo of a problem, get the solution, ask follow-ups. That loop brought massive traffic, but treated every student the same way. Our mission: raise retention, and grow the product into something that helps students far beyond homework.
The turning point
Why change
By 2024 the answer machine was winning, and stalling. Three forces pushed us to rebuild:
Leaky retention
Usage was transactional: come with one problem, copy the answer, leave. Exam weeks spiked; loyalty never formed.
A commoditized loop
Every competitor shipped the same snap-solve flow. Faster answers were no longer a moat.
A new AI window
LLMs had just made one-on-one teaching possible at scale. For us, or for whoever moved first.
Staying an answer machine meant staying replaceable.
The strategy
One Promise, Four Layers
The master plan
This is the blueprint I led the redesign by: one promise, delivered in four layers of product work. The rest of this case follows it in build order — each layer below is a chapter, from 01 Core to 04 Personal.
THE PROMISE
How does a shared answer machine become a school that knows each student?
Our answer to all three forces. Four layers, built from the core up.
04
Personal · Streaks plus proactive agents.
A school that remembers you and reaches out.
03
Reach · Rebuild the IP system.
Change what students think the product is.
02
Stay · Flashcards, study rooms, mock practice.
Turn a quick tool into a long-term study partner.
01
Core · Redesign Ask AI: clearer explanations, more agents, voice.
The front door. If explanations are not clearly better, nothing else matters.
Each layer stands on the one below.
Sharpen the core, build reasons to stay, reach more students, and make it personal.
Layer 01
Sharpening the Core: Ask AI, Explained Better
Why this first
Photo-solving is where every student's journey starts, so it earned the first and biggest investment. The rebuild made two bets: explanations students learn from instead of copying, and a chat calm enough to think in — every extra tool moved out of the way, into a sidebar.
Step-by-step explanations
Before
After
<>
Before — one long block. Easy to copy, hard to learn from.
After — clear steps, answer set apart.
Clean chat + sidebar
Before
After
<>
Before — a toolbar of features inside chat. Distracting, hard to stay in one thread.
After — one clean chat. A curated sidebar: key tools and history, in the student's control.
Layer 02
From Quick Tool to Long-term Study Partner
The shift
Solving one problem is a transaction; learning is a relationship. Exam prep is where the need is sharpest — the hope is that along the way, a little more actually sticks. We cut the features that didn’t earn their place, and kept the methods students already love — backed by data and social buzz.
Flashcards
The method students already know
Mock practice
Walk in ready for the exam
Online study rooms
The flexible slot — focus with company
Layer 03
A New Face: Rebuilding the IP
The perception
Students saw Answer.AI as a vending machine for answers — useful, but not somewhere to stay. The old IP didn’t help: cute but childish, and hard to remember. The new cast grows out of the Answer.AI logo and acts like a school’s favorite teachers — their personalities come through in how each AI tutor talks, so the app starts feeling like a school of their own.
Before
After
<>
Alma — the warm, outdoor-loving counselor.Ace — the playful genius learning buddy.Grant — the shy advisor with a head for numbers.Layer 04
Streaks, Nudges, and a School That Knows You
Make it proactive
The tutors act first. Ace, Alma and Grant get to know each student — grade, school, strong subjects, dream schools — while Ace maps mastery from follow-up questions and scanned problems. Then they reach out at the right moment: a check-in after yesterday’s tough question, a preview invite the night before tomorrow’s calculus class.
The reward loop
A light reward system keeps it going: scanning problems, mastering topics and finishing Ace’s exercises earn badges — plus random outfit drops for an avatar that shows up in the online study room.
Proactive preview
Tomorrow’s class, tonight’s heads-up from Ace.
A call with Alma
Strong in a subject? Ace books a college-planning call.
Badges to avatar
Tasks earn badges — badges dress the study-room avatar.
Honest notes
Learning From a Miss
What went wrong
Not every bet landed. In interviews, students spoke warmly about a counselor, a scholarship advisor and AP prep — and we heard a bigger need than everyday behavior ever showed. Alma, Grant and AP History quietly moved down the roadmap, and a lesson moved up: interviews are polite, behavior is honest. Now the loudest voice in every decision is what students actually do.
The wrap
Grateful, and Onward
Full loop
Design had a seat here from strategy all the way to shipping — rare, and I’m grateful to the team that made it possible ❤️ The app keeps growing on the new foundation. What I’ll remember, though, is meeting students at campus events: pulling up their report cards to show us, saying thanks, telling us it helped. That’s the best part — not the charts, but knowing they feel themselves getting better.
Results
+13%
DAU growth
~1M
New registered users
+23%
Daily revenue
Role
Product Design Lead
Credits
Leo /Product Lead, Hang /Engineer
Time
2025 - Present
+13%
DAU growth
~1M
New registered users
+23%
Daily revenue
The context
Context
Answer.AI is a learning app with nearly 7 million users, most of them middle-school to college students. It grew on one killer loop: snap a photo of a problem, get the solution, ask follow-ups. That loop brought massive traffic, but treated every student the same way. Our mission: raise retention, and grow the product into something that helps students far beyond homework.
The turning point
Why change
By 2024 the answer machine was winning, and stalling. Three forces pushed us to rebuild:
Leaky retention
Usage was transactional: come with one problem, copy the answer, leave. Exam weeks spiked; loyalty never formed.
A commoditized loop
Every competitor shipped the same snap-solve flow. Faster answers were no longer a moat.
A new AI window
LLMs had just made one-on-one teaching possible at scale. For us, or for whoever moved first.
Staying an answer machine meant staying replaceable.
The strategy
One Promise, Four Layers
The master plan
This is the blueprint I led the redesign by: one promise, delivered in four layers of product work. The rest of this case follows it in build order — each layer below is a chapter, from 01 Core to 04 Personal.
THE PROMISE
How does a shared answer machine become a school that knows each student?
Our answer to all three forces. Four layers, built from the core up.
04
Personal · Streaks plus proactive agents.
A school that remembers you and reaches out.
03
Reach · Rebuild the IP system.
Change what students think the product is.
02
Stay · Flashcards, study rooms, mock practice.
Turn a quick tool into a long-term study partner.
01
Core · Redesign Ask AI: clearer explanations, more agents, voice.
The front door. If explanations are not clearly better, nothing else matters.
Each layer stands on the one below.
Sharpen the core, build reasons to stay, reach more students, and make it personal.
Layer 01
Sharpening the Core: Ask AI, Explained Better
Why this first
Photo-solving is where every student's journey starts, so it earned the first and biggest investment. The rebuild made two bets: explanations students learn from instead of copying, and a chat calm enough to think in — every extra tool moved out of the way, into a sidebar.
Step-by-step explanations
Before
After
<>
Before — one long block. Easy to copy, hard to learn from.
After — clear steps, answer set apart.
Clean chat + sidebar
Before
After
<>
Before — a toolbar of features inside chat. Distracting, hard to stay in one thread.
After — one clean chat. A curated sidebar: key tools and history, in the student's control.
Layer 02
From Quick Tool to Long-term Study Partner
The shift
Solving one problem is a transaction; learning is a relationship. Exam prep is where the need is sharpest — the hope is that along the way, a little more actually sticks. We cut the features that didn’t earn their place, and kept the methods students already love — backed by data and social buzz.
Flashcards
The method students already know
Mock practice
Walk in ready for the exam
Online study rooms
The flexible slot — focus with company
Layer 03
A New Face: Rebuilding the IP
The perception
Students saw Answer.AI as a vending machine for answers — useful, but not somewhere to stay. The old IP didn’t help: cute but childish, and hard to remember. The new cast grows out of the Answer.AI logo and acts like a school’s favorite teachers — their personalities come through in how each AI tutor talks, so the app starts feeling like a school of their own.
Before
After
<>
Alma — the warm, outdoor-loving counselor.Ace — the playful genius learning buddy.Grant — the shy advisor with a head for numbers.Layer 04
Streaks, Nudges, and a School That Knows You
Make it proactive
The tutors act first. Ace, Alma and Grant get to know each student — grade, school, strong subjects, dream schools — while Ace maps mastery from follow-up questions and scanned problems. Then they reach out at the right moment: a check-in after yesterday’s tough question, a preview invite the night before tomorrow’s calculus class.
The reward loop
A light reward system keeps it going: scanning problems, mastering topics and finishing Ace’s exercises earn badges — plus random outfit drops for an avatar that shows up in the online study room.
Proactive preview
Tomorrow’s class, tonight’s heads-up from Ace.
A call with Alma
Strong in a subject? Ace books a college-planning call.
Badges to avatar
Tasks earn badges — badges dress the study-room avatar.
Honest notes
Learning From a Miss
What went wrong
Not every bet landed. In interviews, students spoke warmly about a counselor, a scholarship advisor and AP prep — and we heard a bigger need than everyday behavior ever showed. Alma, Grant and AP History quietly moved down the roadmap, and a lesson moved up: interviews are polite, behavior is honest. Now the loudest voice in every decision is what students actually do.
The wrap
Grateful, and Onward
Full loop
Design had a seat here from strategy all the way to shipping — rare, and I’m grateful to the team that made it possible ❤️ The app keeps growing on the new foundation. What I’ll remember, though, is meeting students at campus events: pulling up their report cards to show us, saying thanks, telling us it helped. That’s the best part — not the charts, but knowing they feel themselves getting better.
Results
+13%
DAU growth
~1M
New registered users
+23%
Daily revenue