World Monitor: How Open-Source OSINT Is Democratizing Global Intelligence

When the US and Israel launched strikes against Iran in early 2026, millions of people turned to Twitter, Telegram, and cable news for information. They got a firehose of unverified claims, delayed reporting, and conflicting narratives. Meanwhile, open-source intelligence (OSINT) practitioners were tracking military flight transponders over the Persian Gulf, monitoring AIS signals for naval positioning, watching satellite fire detection for strike impacts, and cross-referencing GPS jamming zones — all in real time, all from publicly available data.

This is what OSINT looks like in 2026. And tools like World Monitor are putting that capability in everyone’s hands — for free.

What Is OSINT, and Why Should You Care?

OSINT — Open Source Intelligence — is the practice of collecting and analyzing information from publicly available sources to produce actionable intelligence. It’s not hacking. It’s not classified. It’s reading what’s already there, but doing it systematically and at scale.

The sources are everywhere:

• ADS-B transponders broadcast aircraft positions. When military tankers circle over the Mediterranean, you know something’s happening before CNN does
• AIS signals track naval vessels. A carrier group repositioning shows up in ship tracking data hours before any official statement
• Satellite imagery from NASA’s FIRMS detects thermal hotspots — fires, explosions, industrial activity — updated every few hours
• RSS feeds from 150+ news sources in 16 languages provide the narrative layer
• Cloudflare Radar exposes internet outages — a country going dark often signals kinetic action
• Conflict databases like ACLED and UCDP track every reported incident with geolocation

What used to require a Bloomberg terminal, a Janes subscription, and a team of analysts ($10K–$50K/year) is now achievable with open data and the right aggregation layer.

World Monitor: The Open-Source Intelligence Dashboard

World Monitor is a TypeScript application that does exactly this aggregation. It pulls from dozens of real-time data sources and renders them on an interactive 3D globe with AI-synthesized intelligence briefs.

Live demos — no account needed:

• worldmonitor.app — geopolitics, military, conflicts, infrastructure
• tech.worldmonitor.app — startups, AI/ML, cloud, cybersecurity
• finance.worldmonitor.app — 92 stock exchanges, 19 financial centers, 13 central banks

All four variants (including a “Happy Monitor” for positive news) run from a single codebase.

A Real Scenario: Monitoring a Conflict in Real-Time

Let’s walk through how World Monitor’s layers work together during a geopolitical crisis.

Step 1: Signal detection. The dashboard’s anomaly detection system uses Welford’s online algorithm to compute streaming statistics over a 90-day window. When military flights over a region spike to 3.2x the normal Thursday baseline, a z-score alert triggers automatically.

Step 2: Convergence. The focal point detection engine notices that this military flight anomaly coincides with naval AIS activity in the same region, internet outages detected by Cloudflare Radar, and a spike in keywords like “strike” and “IRGC” across RSS feeds. Multiple signal types converging geographically gets flagged as a convergence zone with escalated severity.

Step 3: Visualization. On the 3D globe (deck.gl + MapLibre GL JS, WebGL-accelerated), you see:

• Red conflict markers from UCDP/ACLED
• Aircraft icons from ADS-B military flight tracking
• Ship positions from AIS naval monitoring
• Orange dots from NASA FIRMS satellite fire detection
• Purple hexagons showing GPS/GNSS jamming zones from ADS-B transponder analysis
• Internet outage overlays from Cloudflare

Step 4: AI synthesis. The AI layer generates a situation brief using a 4-tier fallback chain — Ollama (local LLM) → Groq → OpenRouter → browser-side T5 via Transformers.js. No single point of failure. The brief correlates entities across all signal types and produces a Country Instability Index (CII) score.

Step 5: Sharing. You generate a country intelligence dossier — CII score, threat counts, theater posture, prediction markets, 7-day timeline — and share it as a deep link or exported image with QR code.

Here’s what a live MENA view looks like with sanctions, military, waterways, outages, and Iran strike layers enabled: World Monitor — MENA theater, 7-day view. Everything is encoded in the URL — zoom level, active layers, time range — making it trivially shareable.

This entire workflow runs in a browser tab. No credentials needed for the hosted version. Self-hosted, you can run the AI entirely on your own hardware with Ollama — no data leaves your machine.

How It Works Under the Hood

The 3D Globe: 40+ Layers at 60fps

The map renders on deck.gl + MapLibre GL JS — the same stack behind Uber’s visualization platform. Thousands of concurrent markers at 60fps via WebGL.

The UX problem with dense data is solved through progressive disclosure:

• At world zoom, markers fade to 0.2 opacity
• Zoom in and they sharpen to full visibility
• Detail layers (military bases, nuclear facilities, datacenters) only appear at appropriate zoom levels
• Overlapping BREAKING badges are deconflicted by priority — highest-severity first

Smart clustering via Supercluster groups nearby markers at low zoom, expanding on zoom-in with adaptive thresholds. Eight regional presets (Global, Americas, Europe, MENA, Asia, Africa, Oceania, Latin America) and time filtering (1h to 7d) let you slice the data quickly.

URL state sharing means every view is a permalink: ?view=mena&zoom=4&layers=conflicts,bases — useful for sharing a specific intelligence picture.

AI: Local-First, Cloud-Optional

The AI design philosophy is what sets this apart from commercial tools. The 4-tier fallback chain means:

1. Ollama/LM Studio — your hardware, your data, zero API keys
2. Groq — fast cloud inference if local isn’t available
3. OpenRouter — broad model access as second cloud fallback
4. Transformers.js — T5 model running entirely in the browser as last resort

Each tier gets a 5-second timeout before falling through. Results are Redis-cached (24h TTL) and content-deduplicated — identical headlines across concurrent users trigger exactly one LLM call.

Beyond summarization:

• Hybrid threat classification — keyword classifier with async LLM override for higher confidence
• Trending keyword spike detection — 2-hour rolling window vs 7-day baseline, with CVE/APT entity extraction
• Country Instability Index — 23 tier-1 nations have tuned risk profiles; all others get universal scoring from available event data

News: Instant First, Smart Later

A two-phase render avoids the classic “ML clustering blocks the UI” problem:

1. News items appear immediately as a flat list when feed data arrives
2. Topic grouping, entity extraction, and sentiment analysis run asynchronously and upgrade the view progressively

Virtual scrolling with element pooling handles long lists. Custom keyword monitors with word-boundary matching let you set up alerts for specific terms — “IRGC” without matching “CIRGCULAR.”

150+ RSS feeds span 7 locales with native-language sources (French loads Le Monde and France24, Turkish loads BBC Türkçe and Hurriyet, etc.). AI translation enables cross-language intelligence gathering.

Live Video: 30+ Streams

8+ default live video streams (Bloomberg, Sky News, Al Jazeera, Euronews, France24, CNBC, Al Arabiya) plus 30+ expandable channels. 22 live webcams from geopolitical hotspots across 5 regions — including a dedicated Iran/Attacks tab with a 2×2 grid of Tehran, Tel Aviv, and Jerusalem feeds for real-time visual monitoring during escalation events.

HLS native streaming for 10 channels bypasses YouTube’s cookie popups and bot checks. Idle-aware playback pauses streams after 5 minutes of inactivity.

The Tech Variant: Why Developers Should Care

The Tech Monitor variant is purpose-built for the tech community:

• Startup and AI/ML news from curated RSS feeds focused on tech, cloud, and cybersecurity
• Tech company HQs — Big Tech, unicorns, and public companies plotted on the globe
• Cloud regions — AWS, Azure, GCP infrastructure mapped
• Startup hubs with funding data
• Accelerators — YC, Techstars, 500 Startups locations
• Upcoming tech conferences
• Cyber threat IOCs — C2 servers, malware hosts, phishing infrastructure geolocated

For developers, the codebase itself is interesting:

• Proto-first API contracts — 20 typed services with auto-generated clients, servers, and OpenAPI docs. Enterprise-grade API design in an open-source project
• Tauri desktop app with Node.js sidecar, OS keychain integration, and cloud fallback
• PWA with offline map support — Service Worker caches map tiles for offline use
• 16-language i18n with lazy-loaded bundles and RTL support

The project is AGPL-3.0 licensed, which means any modifications to the server must be open-sourced — good for the community, important to know if you’re building on top of it.

Finance Variant: Bloomberg for the Rest of Us

The Finance Monitor tracks:

• 92 global stock exchanges with market caps and trading hours
• 19 financial centers ranked by Global Financial Centres Index
• 13 central banks with BIS policy rates and credit-to-GDP ratios
• 64 Gulf FDI investments (Saudi/UAE) plotted globally, color-coded by status
• WTO trade policy — active restrictions, tariff trends, bilateral flows
• 7-signal market radar — crypto, ETF flows, stablecoin pegs, Fear & Greed, BTC technicals, JPY liquidity, QQQ/XLP macro regime → composite BUY/CASH verdict

How It Compares to Commercial Tools

Feature	World Monitor	Janes / Recorded Future	Dataminr
Price	Free	$10K+/year	$20K+/year
Self-hostable	✓	✗	✗
Local LLM	✓	✗	✗
3D globe	✓	Partial	✗
Open source	AGPL-3.0	✗	✗
Desktop app	✓ (Tauri)	Web only	Web only
Financial data	92 exchanges	Limited	✗

The trade-off: commercial tools have proprietary data sources, dedicated analyst teams, and enterprise SLAs. World Monitor aggregates public data with AI synthesis. For individual researchers, journalists, developers, and small teams — the value proposition is hard to beat.

Getting Started

Fastest path: visit worldmonitor.app — no account, no setup.

Desktop app: download from the releases page for macOS, Windows, or Linux.

Local AI: install Ollama, pull a model (ollama pull llama3.1:8b), and point World Monitor’s settings at http://localhost:11434. Models are auto-discovered.

Self-host: clone the repo, configure data source API keys, deploy. Full documentation included.

Why This Matters

OSINT used to be the domain of intelligence agencies and well-funded research firms. Tools like World Monitor are part of a broader shift: the democratization of intelligence analysis. The data was always public — what changed is the aggregation layer, the AI synthesis, and the visualization.

In a world where conflicts unfold on live ADS-B feeds before they hit the news cycle, having the tools to make sense of the signal noise isn’t just interesting — it’s increasingly necessary. Whether you’re a developer curious about the architecture, a researcher tracking geopolitical events, or just someone who wants to understand what’s actually happening in the world, World Monitor is worth exploring.

The full source is at github.com/koala73/worldmonitor.

---

For more on self-hosted AI tools, see my posts on deploying OpenClaw on AWS with Bedrock AgentCore and comparing self-hosted AI assistant frameworks.