Building the Consensus Graph
Wikipedia has a problem.
When you read a Wikipedia article, you're reading what anonymous editors decided to include. You don't see the debates. You don't see the deletions. You don't see the conflicts of interest.
The founder of Wikipedia, Larry Sanger, has called the platform "badly biased" and warned that it can no longer be trusted as a neutral source.
We're building something different.
What is the Consensus Graph?
The Consensus Graph is our answer to the question: "How do we know what's true?"
It's not a repository of facts. Facts are static. The Consensus Graph is dynamic.
It's not a collection of verdicts. Verdicts imply finality. The Consensus Graph evolves.
It's a living, transparent map of evidence—what supports a claim, what contradicts it, and who contributed each piece.
Every claim exists as a node. Every piece of evidence is an edge. Every contributor is visible. Every relationship is auditable.
The Three Principles
1. Transparency Over Authority
Traditional sources ask you to trust them. Trust the journalist. Trust the editor. Trust the institution.
The Consensus Graph doesn't ask for trust. It earns it through radical transparency.
For any claim in the graph, you can see:
- Every piece of evidence supporting it
- Every piece of evidence contradicting it
- Who submitted each piece
- What sources they used
- How the evidence has changed over time
You don't have to take our word for it. You can verify it yourself.
2. Evidence Over Opinion
The internet is full of opinions. Everyone has one. Few are backed by evidence.
The Consensus Graph centers evidence. Not "I think" or "I feel" or "Everyone knows." What sources exist? What do they show? How reliable are they?
Opinions can be expressed in the discussion layer. But the graph itself tracks only verifiable claims with traceable sources.
If you can't cite it, you can't contribute it.
3. Spectrum Over Binary
Truth is rarely black and white. Yet most fact-checking reduces everything to TRUE or FALSE.
The Consensus Graph uses a spectrum:
- Verified: Overwhelming supporting evidence
- Mostly True: Strong evidence with minor caveats
- Mixed Evidence: Valid points on both sides
- Under Review: Insufficient evidence to assess
- Mixed False: Some truth, but misleading overall
- Mostly False: Contains kernels of truth, but fundamentally wrong
- False: Contradicted by verified evidence
This nuance matters. Many claims are "mostly true" or "mixed"—and pretending otherwise is its own form of misinformation.
How It Works
Submitting a Claim
Anyone can submit a claim to the Consensus Graph. The claim gets parsed, structured, and added as a node.
Initially, the claim sits in "Under Review" status. There's no evidence yet. No consensus.
Contributing Evidence
Verifiers contribute evidence—primary sources, secondary reporting, official documents, data sets. Each piece of evidence is categorized:
- Supporting: Tends to confirm the claim
- Contradicting: Tends to refute the claim
- Contextual: Provides background without direct confirmation or refutation
Evidence is weighted by:
- Source credibility
- Contributor reputation
- Primacy (original source vs. derivative)
- Corroboration (confirmed by other sources)
The Consensus Gauge
As evidence accumulates, the Consensus Gauge shifts.
This isn't a vote. It's a weighted aggregation of evidence quality and contributor reputation. A single piece of excellent evidence from a high-reputation Maven can outweigh dozens of weak contributions from new accounts.
The gauge updates in real-time as new evidence arrives. It's never "final"—because new evidence can always emerge.
Auditing the Graph
Every change to the graph is logged. Every contribution is attributed. Every shift in consensus is explained.
If the gauge moves from "Mixed Evidence" to "Mostly True," you can see exactly why. Which evidence tipped the scale? Who contributed it? When?
Nothing is hidden. Everything is auditable.
What Makes This Different from Wikipedia
Attribution
Wikipedia contributions are anonymous. Consensus Graph contributions are attributed to real accounts with public track records.
Versioning
Wikipedia shows the current version. History is buried. The Consensus Graph makes history visible—you can see how understanding evolved over time.
Evidence-Centered
Wikipedia is article-centered. You read prose. The Consensus Graph is evidence-centered. You examine claims and the data supporting them.
Real-Time
Wikipedia is edited by volunteers on their own schedule. The Consensus Graph incorporates new evidence continuously.
Disagreement is Visible
Wikipedia hides edit wars. The Consensus Graph surfaces disagreement. If experts conflict, you see the conflict—not a false consensus produced by whoever was most persistent.
The Graph in Action
Let's walk through an example.
Claim: "Company X's product caused health problems in 2024."
The claim enters the graph. Initially: Under Review.
Day 1: A Verifier submits three news articles reporting the allegations. The articles are derivative—they all cite the same original lawsuit filing.
Evidence is categorized as: Supporting (3 items, low weight due to single-source dependency)
Gauge: Slightly toward "Mixed Evidence"
Day 3: A Maven submits the original court filing, plus Company X's official response denying the claims.
Evidence update: Supporting (+1 primary source), Contradicting (+1 official denial)
Gauge: "Mixed Evidence"
Week 2: An investigative journalist publishes with new documents. Another Verifier submits an independent medical study relevant to the product.
Evidence update: Supporting (+2, including one high-credibility investigation)
Gauge: "Mostly True"
Month 2: Court case resolved. Company X settles but admits no wrongdoing. Another study finds no conclusive link.
Evidence update: Contextual (+1 settlement), Contradicting (+1 study)
Gauge: "Mixed Evidence" with notation about legal settlement
At any point, you can:
- See all evidence that led to the current gauge position
- See who contributed each piece
- See how the gauge has moved over time
- Form your own judgment based on the underlying evidence
Limitations
The Consensus Graph is powerful, but we're honest about its constraints.
Coverage is not universal. We start with claims that Verifiers submit. Popular topics get more attention than obscure ones.
Speed varies. Breaking news may not have sufficient evidence for assessment. We won't pretend to know things we don't know.
Manipulation is possible. Sophisticated actors could attempt to flood the graph with coordinated low-quality evidence. We have defenses, but no defense is perfect.
Interpretation requires judgment. The graph shows evidence. What that evidence means can still be debated. We provide the map; you still have to navigate.
The Future
The Consensus Graph is infrastructure. Like any infrastructure, it improves over time.
We're building:
- Cross-claim connections: How does evidence for Claim A affect Claim B?
- Temporal tracking: How do narratives evolve over weeks and months?
- Source networks: Which sources cite which others?
- Predictive signals: What early indicators suggest a claim will shift?
- API access: For researchers, journalists, and enterprise partners
The goal is simple: make truth accessible.
Not truth as someone else defines it. Truth as the evidence supports it.
Join the Build
The Consensus Graph isn't something we build for you. It's something we build with you.
Every piece of evidence you contribute. Every source you trace. Every analysis you provide.
You're not just using the platform. You're building the infrastructure of trust for the next generation.
Be curious again.