OneSource is a project for studying internet attention patterns.
Right now it focuses on ecommerce creatives — things like hooks, emotions, visuals, CTAs, platforms, and niches — and tries to organize them into signals that can actually be explored and compared.
The goal isn’t to generate more AI content.
The goal is to understand why certain creative patterns keep working.
clean signal > more AI
Most tools try to help people create more content faster.
OneSource goes in the opposite direction.
It tries to slow things down and study the structure behind attention:
- what hooks repeat
- which emotions appear together
- how visuals reinforce persuasion
- what changes across platforms
- which patterns keep showing up again and again
It started with ecommerce because ecommerce creatives are structured enough to compare reliably.
But the bigger idea is broader than ecommerce.
OneSource is not:
- an AI wrapper
- a chatbot
- a content generator
- a viral prediction tool
- a recommendation engine
AI is useful during development and organization, but it’s not really the core idea here.
The interesting part is the structure:
- normalized signals
- repeatable patterns
- relationships between creative behaviors
- accumulated attention data over time
The system currently follows a pretty simple flow:
creative data
→ signal cleanup
→ pattern extraction
→ exploration workspace
Each creative can contain signals like:
- emotions
- hooks
- visual styles
- CTAs
- platforms
- niches
From there, OneSource looks for:
- repeated patterns
- strongest co-signals
- reinforced structures
- platform behavior
- signal relationships
The idea is to make attention patterns easier to inspect instead of guessing what “might go viral.”
OneSource is currently in what I call the “Pause Phase.”
The project was evolving quickly, so I intentionally stopped adding major features for a while and focused on making the foundation cleaner and easier to understand.
Most of the recent work has been around:
- stabilizing the intelligence layer
- cleaning signal behavior
- improving exploration UX
- fixing inconsistent traversal logic
- simplifying the product story
- reducing unnecessary complexity
That restraint ended up helping the product a lot.
Current foundations include:
- governed signal vocabulary
- signal quality validation
- normalized signal extraction
- universal signal exploration
- weighted signal rendering
- breadcrumb traversal
- strongest co-signal analysis
- pattern + platform intelligence
- minimal Vitest coverage
- production build + lint verification
The app now feels more like an intelligence workspace than a tagged gallery.
The Next.js app lives in:
onesource/
Main routes:
/→ entry screen/gallery→ intelligence workspace/admin→ creative ingestion + governance
Commands:
cd onesource
npm run dev
npm run test
npm run lint
npm run buildCurrent stack:
- Next.js
- TypeScript
- Tailwind
- Supabase
- Vitest
Things intentionally NOT being added yet:
- graph systems
- AI reasoning engines
- recommendation systems
- embeddings/vector infrastructure
- ontology expansion
- feature-heavy UX
Trying to keep the foundation clean before making the system more complex.
The long-term idea is something closer to:
a structured intelligence layer for internet attention
or:
a creative genome for persuasion patterns
Most products help generate more content.
OneSource is more interested in understanding the patterns underneath it.