AI-assisted coding workflow
Build Notes
Product Build Notes
Development records covering launch decisions, AI-assisted workflows, store publishing, UI experiments, and what changed between iterations.
Play Store publishing notes
UI/UX iteration logs
Product journey writeups
LIVE / Language Learning
Hallo German
Designed for lightweight practical German learning.
Why native Android
The product needed smooth daily learning screens, offline-friendly progress, and Android-specific language input flows, so it was built as a Kotlin and Jetpack Compose app instead of a generic cross-platform prototype.
Learning data structure
Words and sentences are organized by level and topic, then connected to flashcards, listening, reading, writing, speaking-oriented practice, notebook review, and learning-history screens.
Device features
Room and DataStore keep local study state stable, while ML Kit dependencies prepare OCR, translation, and handwriting recognition paths for learners who want to capture study material from the real world.
LIVE / Utility
Lotto Sago Sipeunnal
Built around fast repeat use and clear number generation.
Beyond a generator
The app started from quick number generation, but the product shape became stronger when saved sets, draw results, QR scan history, reminders, and winning-store context were treated as one weekly lotto workflow.
Local-first records
Hive boxes are used for generated sets, result history, scan history, and store data so users can return to previous combinations and checks without depending on a heavy account system.
Release constraint
Store publishing exposed a practical country-availability issue: Korea is a future launch target, but it is currently excluded because of app rating constraints that need to be resolved first.
LIVE / File Transfer
SwiftBeam: Any Device Transfer
Explores frictionless local sharing across phones and desktops.
Multiple transfer routes
The hard part was not a single fast path, but deciding what should happen when two devices cannot use the same path. The coordinator chooses between Nearby, Wi-Fi Direct, LAN TCP, FTP, web portal, and manual fallback routes.
Mixed-device reality
Phones and desktops have different permissions and network behavior, so QR connection, receiver confirmation, progress states, overwrite/skip/cancel handling, and browser fallback are treated as first-class product flows.
Release packaging
Mobile store releases and desktop downloads were prepared together, which made versioned Windows installer, portable package, macOS archive, and store badges part of the same product delivery story.
BETA / AI Productivity
QR Scan4u
A camera-first utility refined through quick workflow experiments.
Camera-first utility
The prototype focuses on the moment after scanning, not only on detection speed: the result needs to be readable, trustworthy, and immediately actionable.
Device capability baseline
The build is intentionally small so camera permission, recognition state, result parsing, link handling, and copied-data flows can be tested without a heavy product shell.
Next product question
The remaining work is deciding whether this should stay a compact scanner or grow into a history, batch action, and AI-assisted interpretation tool.
ACTIVE / AI Productivity
AI Blogger
Turns AI-assisted blog writing into a controlled desktop publishing workflow.
From Tistory tool to AI Blogger
The product name and scope changed because the workflow expanded from Tistory-only publishing into a broader desktop system that can manage both Tistory and Naver blog operations.
Desktop plus backend split
Electron owns the dashboard, settings, preview, licensing, and packaged app flow, while the Python backend handles browser automation, platform login, publishing commands, and AI-assisted generation work.
Operational safeguards
Local JSON stores, SafeStorage, session files, preview mode, publishing history, and category templates were added because blog automation needs reviewability and recovery, not just one-click generation.
PROTOTYPE / Game Prototype
Puzzle Game
An interaction study around tactile puzzle assembly.
Implementation note
Custom interaction logic
Implementation note
Animation tuning for responsiveness