FRICTION

"Users do NOT think while confused."

FRICTION is a minimalist AI-powered productivity ecosystem designed to bridge the gap between raw information capture and meaningful insights.

Instead of interrupting your workflow with AI suggestions, FRICTION captures your moments of confusion, curiosity, and realization instantly. AI analyzes them later in batches, helping you discover learning patterns you would otherwise miss.

---

✨ Product Philosophy

⚡ Capture Instantly, Analyze Later

Capture thoughts without breaking focus. Reflection happens later.

🤫 Systemic Silence

AI acts as a silent analyst—not a tutor, chatbot, or distraction.

📈 Patterns Over Moments

One moment rarely matters. Meaning emerges from hundreds of moments over time.

🧠 Preserve Context

Raw data should remain untouched until reflection.

Raw Data ≠ Interpreted Data

---

🚀 Features

• ⚡ Instant keyboard-driven capture
• 📝 Capture highlights, notes, and learning moments
• 🤖 AI-generated learning insights
• 🔍 Detect recurring confusions and knowledge gaps
• 📊 Weekly & long-term learning pattern analysis
• 🧠 Persistent Learning Records
• 📚 Searchable personal knowledge base
• 🔒 Google OAuth authentication
• 🎯 Zero-friction workflow with delayed intelligence

---

🏗️ Architecture

FRICTION follows a simple Capture → Process → Review lifecycle.

Chrome Extension
        │
        ▼
 Capture Raw Moments
        │
        ▼
 Backend Snapshot Service
        │
        ▼
 Gemini Batch Analysis
        │
        ▼
 Candidate Findings
        │
        ▼
 Review Dashboard
        │
        ▼
 Stable Learning Records

---

🧩 Components

Component	Responsibility	Tech Stack
Chrome Extension	Keyboard-driven raw capture	Manifest V3
Backend API	Authentication, orchestration, snapshot pipeline	Node.js, Express
Web Dashboard	Review findings and visualize trends	React, Tailwind CSS
Database	Store moments, findings, learning records	MySQL
AI Layer	Batch analysis and pattern detection	Gemini

---

📦 Data Flow

Raw Moments
      │
      ▼
Snapshot Created
      │
      ▼
Gemini Batch Analysis
      │
      ▼
Candidate Findings
      │
      ▼
User Review
      │
      ├── Confirm
      ├── Defer
      └── Reject
      │
      ▼
Learning Records

---

🗄️ Data Models

Buffer Moments

moment_id      CHAR(36)
user_id        CHAR(36)
raw_text       TEXT
source_type    ENUM('highlight','bulk_paste')
source_url     TEXT
status         ENUM('pending','processed')
created_at     TIMESTAMP

---

Candidate Findings

finding_id            CHAR(36)
snapshot_id           CHAR(36)
type                  ENUM(
    'confusion',
    'insight',
    'fragile_understanding',
    'pattern'
)
topic                 VARCHAR(255)
evidence_moment_ids   JSON
state                 ENUM(
    'unreviewed',
    'confirmed',
    'deferred',
    'rejected'
)

---

🔄 State Transitions

Buffer Moments

pending
    │
    ▼
processed
    │
    ▼
deleted

---

Findings

unreviewed
      │
      ├────────► confirmed
      │
      ├────────► deferred
      │
      └────────► rejected

---

Learning Records

Learning Records are created or updated only after a finding has been:

• ✅ Confirmed
• ⏳ Deferred

Rejected findings are permanently discarded.

---

🌐 API

Capture

POST /api/moments

Capture raw text with metadata.

---

Snapshot

POST /api/snapshots/run

Trigger batch AI analysis.

Maximum:

• 30 moments per snapshot

---

Review

GET /api/findings?state=unreviewed

Fetch pending findings.

POST /api/findings/:id/confirm

Confirm a finding.

DELETE /api/findings/:id

Reject a finding permanently.

---

⚙️ Tech Stack

Frontend

• React
• Tailwind CSS

Backend

• Node.js
• Express

AI

• Gemini

Database

• MySQL

Authentication

• Google OAuth

Browser

• Chrome Extension (Manifest V3)

---

🎯 Version 1 Constraints

• ❌ No RAG
• ❌ No vector database
• ❌ No auto-teaching
• ❌ No AI chat
• ✅ Batch-only AI analysis
• ✅ Google OAuth authentication
• ✅ Human-in-the-loop validation

---

📌 Vision

Most note-taking apps remember what you learned.

FRICTION remembers how you learn.

Instead of storing information, it uncovers recurring mistakes, fragile understanding, and hidden learning patterns—helping you build lasting knowledge over time.

---

📈 Success Metric

"This surfaced a pattern I didn't realize."