Major additions: - Community Next.js app (port 18187) for browsing claims with API docs - stemedb-chaos crate: Fault injection, chaos testing, CRDT properties - Latent ingestion system: Reddit/FDA ingesters with ADK-Go agents - Disputed claims handling: Manual review workflows and validation - Aphoria security scanner: New extractors (SQL injection, command injection, weak crypto, TLS version), policy-based ignores, UAT reports - Docker infrastructure: Dockerfile, docker-compose.yml for full stack - VulnBank demo: Intentionally vulnerable multi-language test corpus SDK & API enhancements: - Source registry handlers for tracking data provenance - Metrics endpoint - Skeptic filtering improvements Code quality: - Split 14 large files (>500 lines) into focused modules - All files now under 500-line limit per project guidelines Documentation: - Chaos testing guide, circuit breakers, observability docs - Phase 7 UAT documentation updates - Martin Kleppmann technical writer agent Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
105 lines
4.0 KiB
Markdown
105 lines
4.0 KiB
Markdown
# 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.*
|