← All Projects
Meridian · Epistemic Infrastructure

Cronkite

Epistemic scoring for the modern information environment. Because trust has to be earned, measured, and maintained.

Active · v0.8 · Chrome Extension + API
Consensus
87
Valid
88
Reliable
92
Value
85
Transparency
83
Scored as News article — CLAUDE + GROK + GPT + GEMINI + MISTRAL
Register: FACTUAL · Provenance: Primary Source 94
Overview

The information environment is degraded not because truth doesn't exist — but because no tool has measured epistemic quality at scale, in real time, for everyone.

Request API Access

The ambient immune system
the internet never had.

Cronkite scores any web content — news articles, social media posts, government documents, research papers, blog posts — across five epistemic dimensions using a five-model AI ensemble. Five independent models evaluate independently. Their scores are aggregated using a median, with disagreements surfaced and displayed. No single model dominates. No single bias prevails.

A crowdsourced domain trust database builds itself passively through normal browsing. The first person to read an article scores it for everyone who follows. By the time content goes viral, the epistemic signal is already cached and ready to serve at zero marginal cost to the next reader.

Cronkite addresses what global risk frameworks have consistently ranked at or near the top of the risk landscape two years running: misinformation and disinformation. Not by censoring, not by flagging, not by deciding what is true — but by making the epistemic quality of information visible, interpretable, and available to everyone.

How It Works

Five steps.
One signal.

1
Page detected
Chrome extension identifies the URL and extracts clean text from the DOM, skipping navigation and noise.
2
Source classified
AI classifies the source type and rhetorical register — FACTUAL, MIXED, PERFORMATIVE, ACTIVATING, or HISTORICAL.
3
Five models score
Claude, Grok, GPT, Gemini, and Mistral evaluate independently across five epistemic dimensions in parallel.
4
Median aggregated
Scores are aggregated via median. Spread above 20 points triggers a Highly Contested flag. Disagreements are surfaced.
5
Score cached
Result is stored in the domain trust database. Every subsequent reader gets the result instantly at zero cost.
Epistemic Dimensions

Five dimensions.
One score.

Dimension 01
Valid
Are the claims supported by evidence within the text? Cronkite evaluates sourcing structure — named sources, direct quotes, cited reports — not factual accuracy against external knowledge. The question is whether evidence is present and internally consistent, not whether it matches a model's training data.
Dimension 02
Reliable
Could someone follow the same process and reach similar conclusions? Reliability measures the traceability of claims — whether the methodology is described, sources are named, and the path from evidence to conclusion is visible. For social media, it evaluates whether a speaker's record is consistent with their current claims.
Dimension 03
Value
Does this content help someone make a better-informed decision? Value measures epistemic utility — the degree to which the content adds genuine information rather than emotional activation. A papal statement on poverty scores differently from a conspiracy claim, because their purpose and informational contribution are structurally different.
Dimension 04
Transparency
Is the source honest about who made it and why? Transparency evaluates author disclosure, institutional affiliation, conflict of interest disclosure, and funding transparency. A think tank brief that doesn't disclose its funder scores lower than one that does, regardless of content quality.
Dimension 05 — Provenance (displayed separately)
Originality
Where does this content sit in the information chain? Originality is scored across four tiers: raw primary source (95-100), original investigative journalism (88-94), professional secondary analysis (65-87), minor amplification (35-64), and reactionary or anonymous content (0-34). A CNN exclusive investigation scores Tier 1. Fox News covering CNN's story scores Tier 2. Provenance is displayed separately from the overall score — it is a structural fact about the information chain, not an epistemic quality judgment.
The Register System

Context-aware
epistemic standards.

Cronkite's most important innovation: no other epistemic scoring system asks what kind of content it is evaluating before applying standards. A papal declaration and a Reuters news article should not be scored by the same rules. Cronkite classifies the rhetorical register first — then applies appropriate epistemic standards.

Factual
Primary purpose is to inform with verifiable claims. Full evidentiary standards applied.
Reuters article on inflation data. NEJM clinical trial. Court ruling.
Mixed
Factual claims in promotional or partisan framing. Evaluate data, not spin.
Earnings call. Ideologically-leaning think tank brief. Advocacy op-ed.
Performative
Relational, ceremonial, or expressive. Not making falsifiable claims. Asking for sources would be rude.
Macron's Vatican post. Presidential inauguration address. Papal blessing.
Activating
Falsifiable claim weaponized as emotional trigger. Asking for sources is appropriate.
"I WON THE ELECTION BY A LOT!" Conspiracy content. Outrage bait.
Historical
Primary historical documents. Era-appropriate standards. Not modern truth claims.
The Bible. The Magna Carta. Common Sense. The Federalist Papers.
Named For

"And that's the way it is."

Walter Cronkite, CBS Evening News

Walter Cronkite anchored the CBS Evening News from 1962 to 1981. He was named the most trusted man in America in a 1972 poll — not because he was infallible, but because he was consistent, transparent about what he knew and didn't know, and committed to the separation of fact from opinion.

When Cronkite returned from Vietnam in 1968 and told Americans on air that the war was a stalemate, President Johnson reportedly said: "If I've lost Cronkite, I've lost Middle America." That is epistemic authority. Not popularity, not charisma — the earned trust of an audience that knew he would tell them what was true even when it was inconvenient.

Cronkite the product is named for that standard. Not a fact-checker. Not a censor. An instrument for making epistemic quality visible — so that trust, like Cronkite's, can be earned rather than assumed.

Try Cronkite.
Free. Always.

The Chrome extension is free. The score is free. Epistemic access is not a premium feature. API access for institutions, researchers, and developers available on request.

At a Glance
Status Active · v0.8
Models 5-model ensemble
Cost per score $0.02 – $0.04
Platforms Chrome · API
Risk Domain #1 — Misinformation
Also from Meridian

The full portfolio.

Open-Source Threat Surveillance
Snow
AI-driven monitoring of global news and media feeds to detect infectious disease outbreaks, CBRN incidents, and humanitarian crises before institutional response becomes reactive.
Country Risk Intelligence
Marshall
Living country risk scores across a structured global risk taxonomy — updated continuously, built for policy makers, NGOs, researchers, and institutional decision-makers.

Meridian is the independent research and applied-technology project of Mark Greenhalgh. Views, analysis, and materials published under the Meridian name are his own and do not represent the position of the U.S. Department of Defense, the Department of the Army, or any U.S. government agency or component. Meridian operates independently of his official duties.