How pulsEva Analyzes Intelligence
pulsEva ingests intelligence from multiple curated sources across wire services, broadsheets, regional media, think tanks, government publications, and social channels. Every article passes through a multi-stage NLP pipeline that classifies content along three independent axes, extracts named entities, geocodes locations, and assigns NATO-standard source grades before reaching your feed.
The pipeline runs on 10-minute polling cycles. Content is deduplicated by SHA-256 hash, enriched with spaCy NER and Gemini-powered classification, and clustered into story arcs by embedding similarity. The result is an intelligence feed where every item carries provenance, confidence, and analytical context.
The PSESII Framework
Every piece of evidence is classified into one or more of six analytical domains. The PSESII framework is the backbone of intelligence structuring used by NATO-aligned analytical organizations.
Political (P)
Elections, diplomacy, sanctions, protests, governance transitions, and legislative developments that shape the geopolitical landscape.
Security (S)
Armed conflict, military posture, defense spending, NATO alliances, terrorism, organized crime, and state-level cyber operations.
Economic (E)
Trade policy, financial markets, energy commodities, supply chain disruptions, currency movements, and corporate developments.
Social (So)
Migration flows, public health emergencies, human rights developments, demographic shifts, and humanitarian crises.
Information (I)
Disinformation campaigns, media freedom, civilian cyber security, technology and AI advances, espionage, and information control.
Infrastructure (In)
Critical infrastructure, transportation networks, energy grids, telecommunications, water systems, and natural disaster impacts.
Source Grading (NATO Admiralty Code)
The NATO Admiralty Code grades two different things. Reliability (A-F) measures a source's track record and attaches to the source. Credibility (1-6) judges one specific piece of information and attaches to the claim, not to whoever published it. Read together they form a composite like "B2" -- a usually reliable source reporting probably true information.
Today pulsEva assigns the reliability axis only. Every source carries an A-F grade. The credibility digit is the standard we grade information against, but pulsEva does not yet assign it automatically -- so composite grades like "B2", which need both axes, are not displayed. The 1-6 table below is the reference, not a description of something we currently emit.
What you do see alongside a story is a corroboration signal: how many distinct sources carried it, how many of those are state-controlled, and a "reported" or "corroborated" label derived from how many articles the story attracted. That is a volume-and-ownership signal, not an independence assessment -- syndicated republication of a single wire report raises the article count without adding a second independent account of events. Judging independence, and turning it into a credibility digit, is work in progress.
Source Reliability (A-F)
| Grade | Label | Examples |
|---|---|---|
| A | Completely Reliable | Reuters, AP, AFP |
| B | Usually Reliable | BBC, FT, NYT, Al Jazeera |
| C | Fairly Reliable | IISS, RAND, government press |
| D | Not Usually Reliable | Unvetted blogs, tabloids |
| E | Unreliable | Known propaganda outlets |
| F | Cannot Be Judged | Anonymous sources, new accounts |
Information Credibility (1-6)
| Rating | Label | Criteria |
|---|---|---|
| 1 | Confirmed | Confirmed by independent sources |
| 2 | Probably True | Plausible, consistent with known facts |
| 3 | Possibly True | May be accurate, lacks confirmation |
| 4 | Doubtful | More likely false than true |
| 5 | Improbable | Contradicts established facts |
| 6 | Cannot Be Judged | Insufficient information |
The Source Review BoardRolling out
A flat aggregator treats every feed as equally trustworthy. pulsEva does not. Every source carries a NATO Admiralty reliability grade (A-F, detailed above) today; the credibility digit is not yet assigned. The review board that governs how those grades change -- editorial approval with a recorded rationale, written to an immutable log -- is rolling out and is not yet enabled for all accounts. Everything described in this section is built; ask us for current availability before relying on it in an accredited process.
How a grade is reviewed
Eva, our analytical engine, produces a monthly scorecard for each source -- measuring publishing yield, topic and geographic coverage, behavioral drift against its current grade, and how often its reporting is corroborated by independent sources. A human editor then approves, adjusts, or rejects the recommendation. A grade only changes through that review, and every change is written to a permanent log with the reason.
Who controls the source
pulsEva records ownership and state-control -- independent, public-service, state-controlled, state-affiliated, or private-commercial -- and surfaces it. Listing a state broadcaster next to an independent wire service without that context hides the single most important fact about a claim.
Once a source has been through the board, its grading shows a "Reviewed · {date}" stamp with the date it last passed editorial review. Sources that have not been reviewed yet show no stamp -- which is itself the honest answer, and today that is most of them. The stamp is the signal that a grade is no longer a black box, not a claim that none of them are.
Three-Axis Classification
Every piece of evidence is classified along three independent axes. This multi-dimensional approach allows analysts to filter and cross-reference intelligence from any angle.
Geography
ISO 3166 country codes, administrative regions, and geocoded coordinates. Both event location and relevant geographies are tracked -- a conflict between two nations is relevant to both.
PSESII Domain
The analytical framework described above. Each article receives 1-3 PSESII classifications with confidence scores, enabling multi-domain analysis of complex events.
Consumer Topics
11 user-facing topic categories derived from PSESII domains: Conflict and Security, Cybersecurity, Geopolitical Risk, Political Instability, Economy and Trade, Energy, Technology, Climate, Health, Migration, and Organized Crime.
How Eva Works
Eva is pulsEva's AI intelligence analyst. She generates structured briefings, not casual summaries. Her outputs follow intelligence community standards: lead with the most critical signal, quantify when possible, cite every source.
Briefing Types
- Morning Brief -- Daily digest of overnight developments, ranked by salience
- Alert Brief -- Triggered by critical monitor matches, delivered immediately
- Weekly Digest -- Trends, entity frequency analysis, emerging topics
- Country Dossier -- In-depth per-country intelligence assessment (Guru tier)
Generation Process
- 1. Salience scoring -- Evidence ranked by recency, source grade, corroboration count, and user monitor relevance
- 2. Filter and select -- Top evidence selected per user preferences and tier
- 3. Synthesize -- Gemini-powered structured generation with PSESII framing
- 4. Cite and verify -- Every claim attributed to a source with timestamp and grade
Our Sources
pulsEva curates multiple sources across six categories. Each source is assigned a NATO reliability grade and monitored for availability via circuit breakers. Sources are polled on 10-minute cycles with SHA-256 deduplication.
Wire Services
Reuters, AP, AFP, Interfax, Xinhua
Broadsheets
BBC, FT, NYT, Guardian, Le Monde
Regional Media
Local outlets across many countries
Think Tanks
IISS, RAND, Chatham House, CSIS
Government
MoD press, CERT advisories, foreign affairs
Social / OSINT
Verified X accounts, Telegram channels
Ready to start analyzing?
Get access to pulsEva's intelligence feed with NATO-grade source evaluation and AI-powered briefings.