Home

↑

An AI-first bridge for passive language learning.

View project

↑

↑

kat
ya.

back

kat
ya.

Recu

Language Learning App | A.I. First Design Process

3 Months

Passion Project

Quick capture and langauge retention app to help users capture new phrases and words on the go to empower language learning when short for time.

Quick capture and langauge retention app to help users capture new phrases and words on the go to empower language learning when short for time.

THE PROBLEM.

add phrase via shortcut

added to home via MVP web-app

hi-fi WIP designs

Try via Expo Go

↑

the problem.

the problem.

Many people learn languages through audio courses during their morning commutes. However, by the time they arrive at work, the new phrases they’ve heard are often forgotten before they have a chance to write them down or practice.

"I found that the 30-second friction of stopping my bike, unlocking my phone, and typing a note resulted in a 90% drop-off in captured phrases."

THE PROBLEM.

solution.

solution.

The final product is a zero-UI entry point (the Button) paired with a high-utility review dashboard. I reduced capture time from 30 seconds (unlocking, opening notes, typing) to 2 seconds (press and talk).

The final product is a zero-UI entry point (the Button) paired with a high-utility review dashboard. I reduced capture time from 30 seconds (unlocking, opening notes, typing) to 2 seconds (press and talk).

process pt1.

process pt1.

I used Lovable to deploy a functional web app MVP that I could test in minutes, then iterated on it through prompting.

I used Lovable to deploy a functional web app MVP that I could test in minutes, then iterated on it through prompting.

process pt2.

process pt2.

Once I was happy with the flow, I created hi-fi components in Figma and then translated them into a code-first design system in Replit using Swift.

Once I was happy with the flow, I created hi-fi components in Figma and then translated them into a code-first design system in Replit using Swift.

INSIGHT 2.

bg-base

#1C1C1C

bg-surface

#262626

border-strong

#555555

border-subtle

#3A3A3A

text-secondary

#A1A1A1

text-primary

#F9FAFB

primary-default

#FF6F61

primary-hover

#FF8B82 90%

PROCESS PT1.

light

dark

insight 1.

insight 1.

Users often misspoke, or the audio picked up the wrong phrase. I embedded an LLM to help refine the phrase and show users what they got wrong.

Users often misspoke, or the audio picked up the wrong phrase. I embedded an LLM to help refine the phrase and show users what they got wrong.

insight 2.

insight 2.

Users reviewed their captured phrases in the evening. Consequently, I developed the UI with a dark-mode-first approach to reduce eye strain during those sessions.

Users reviewed their captured phrases in the evening. Consequently, I developed the UI with a dark-mode-first approach to reduce eye strain during those sessions.

TAKEAWAYS.
takeaways.

takeaways.

By starting with a working MVP prototype - combining a hardware trigger with a functional AI logic gate - I avoided over-designing features that didn’t solve the core problem: capturing phrases quickly.


This allowed me to iterate rapidly on feedback and deploy updates in minutes. The app is still a work in progress.

By starting with a working MVP prototype - combining a hardware trigger with a functional AI logic gate - I avoided over-designing features that didn’t solve the core problem: capturing phrases quickly.


This allowed me to iterate rapidly on feedback and deploy updates in minutes. The app is still a work in progress.

Let's cooperate

katyajf@gmail.com

light

dark

Work

Blog

Contact

LinkedIn

↑

Substack

↑

katyajf@gmail.com

light

dark