stemedb/applications/disputed/vision.md

Disputed

Your knowledge graph, always listening.

The Problem

You consume hours of podcasts, articles, videos, conversations. Someone says "studies show X helps Y" and you nod along. Three weeks later, someone else says the opposite. You don't notice the contradiction because your brain doesn't have a database.

Information flows through you. You retain fragments. Contradictions slip past unnoticed. You form beliefs without knowing where they came from or what conflicts with them.

The Solution

A desktop app that sits in your menubar. When audio plays or you highlight text, it:

1. Captures - Transcribes audio, reads highlighted text
2. Extracts - Identifies claims ("X causes Y", "studies show...", "A is better than B")
3. Checks - Compares against everything you've ever saved
4. Surfaces - Shows matches, contradictions, and new information

You're not taking notes. You're building a knowledge graph by just living.

How It Feels

You're listening to a podcast. The host says "cold exposure increases dopamine by 250%."

A toast appears:

┌─────────────────────────────────────────┐
│ ✦ NEW CLAIM                             │
│ "cold exposure" → "increases dopamine"  │
│ Source: Huberman Lab #142               │
│ Confidence: 0.8 (cites study)           │
│                                         │
│ ⚠️ Related: 2 saved claims about        │
│    dopamine mention different numbers   │
│                                         │
│ [Save] [Dismiss] [Explore]              │
└─────────────────────────────────────────┘

You tap "Explore" and see the full picture: what you've saved, what agrees, what conflicts, where each claim came from.

Later you're reading an article. You highlight a paragraph. Same flow - claims extracted, checked against your graph, saved if you want.

Core Principles

1. Zero Friction

You don't change your behavior. Listen to podcasts like normal. Read articles like normal. Disputed works in the background, surfacing insights only when they matter.

2. Claims, Not Facts

Disputed doesn't tell you what's true. It shows you the landscape of claims - who said what, when, with what confidence. Truth emerges from seeing the full picture.

3. Your Graph, Your Sources

You control what goes in. Disputed learns what YOU'VE consumed, not some global dataset. Your knowledge graph reflects your information diet.

4. Contradictions Are Features

When sources disagree, that's valuable information. Disputed surfaces disagreement explicitly rather than hiding it behind a single "answer."

Use Cases

Health & Wellness

• Track supplement claims across podcasts, studies, forums
• See when a new recommendation contradicts previous advice
• Know which claims are well-supported vs. single-source

Professional Learning

• Capture insights from conference talks, webinars, courses
• Build institutional knowledge from meetings and calls
• Notice when team members make contradictory claims

Research

• Extract claims from papers automatically
• Track where the literature agrees vs. fights
• See how consensus shifts over time

Media Literacy

• Notice when news sources contradict each other
• Track claims politicians make over time
• See your own information bubble

Technical Foundation

Disputed is built on Episteme (StemeDB), an append-only knowledge graph that:

• Stores claims without forcing resolution
• Handles contradictions as first-class data
• Resolves at read-time through configurable "Lenses"
• Supports time-travel queries ("what did I believe last month?")
• Tracks source provenance and trust tiers

This isn't a wrapper around a vector database. It's a proper epistemological data model.

The Name

Disputed - because the most valuable information is often contested. We don't hide disagreement; we surface it.

---

Your brain doesn't have version control. Now it does.