
KNOWLEDGE APP PET PROJECT
Remember what you
find interesting
3 weeks
Solo Designer
Figma, Claude Code
Coming in a ~week :)




{/1}
PRODUCT FRAME
A knowledge app designed around capturing, organizing, and resurfacing what you'd otherwise forget.
Eidos started as a personal pain point: I kept saving collected things across platforms, but I almost never went back to them. Facts, quotes, articles clips, screenshots scattered across Telegram, Notion, Day One, Notes — everything got stored in different places, and eventually just turned into digital dump. I could not surface something I once found valuable to save, and eventually I just forgot what I had saved.
That’s when I decided that I needed a single platform that would not only store the collected items in neatly organized folders, but would actually bring me back to the things I once saved and help me refresh them in my memory.

One saved thing at a time. AI handles the tagging — you just capture and browse.
{/2}
AI DESIGN PROCESS
Designed in Figma. Built by Claude Code. Guided by a spec I owned end-to-end.
Before any screens existed, I used Claude to figure out what the app should actually be by challenging each other’s assumptions, and considering several approaches to the problem. Once that was settled, an AGENTS.md file and a master spec kept every phase working from the same rules.
Phase 1
Phase 2
Phase 3
Phase 4
Phase 5
Research, scope, PRD
IA: UX & flows
Design (hi-fi mockups) + Design system
Build (frontend, backend,
AI API integration)
Deploy, release
Claude
Claude
Figma + Claude
Claude Code
EAS Build, Google Play
{/3}
END-TO-END FLOW
The full experience — onboarding, first capture & resurfacing — running as a native app on a Pixel 6.
Built with Claude Code. Brought to life with Expo + React Native + Supabase + Gemini.
{/4}
KEY LEARNINGS
One phase at a time
I never asked for the whole app at once. Each phase had to work and get reviewed against the spec before the next one started. That’s the core of AI-native design approach.
Saying No on purpose
The list of features I scoped and chose not to build is longer than what shipped — semantic search, custom folders, tag editing. Staying small enough to ship was the point.
Built for one user, yet
Eidos is organized the way I think, and reflects my habits. Will it serve for other users with different thinking models is the app’s main risk and the next thing to test.