Intelligence Hub Radar
Less noise. More signal. An institutional radar that detects, clusters, scores and explains actionable megatrends — from global information noise to signals with strategic gravity.
What it is not
- · Not an aggregator of "all the news".
- · Not a generic AI newsletter.
- · Not a podcast.
- · Not a self-serve SaaS.
- · Not a full B2B dashboard.
- · Not a mobile app.
- · Not a cheap B2C product.
- · Not a chatbot.
Methodology · Semantic Gravity Engine
- 1. Nightly ingestion. Every night we sweep about a hundred curated sources (primary research, institutional analysis, reference press and curated newsletters).
- 2. Draconian filter. A classifier discards ~90%: listicles, PR, hype, minor increments. A Tier-1 clause protects anything involving globally significant actors.
- 3. Semantic clustering. Articles are grouped by meaning (embeddings + HDBSCAN), not by keywords — one story gathers all its coverage.
- 4. Gravity judge. Each story gets a 0-100 score with explicit reasoning (the one you see in "Why this gravity"). The bias is conservative: we would rather stay silent than inflate.
- 5. Calibration. The published score is calibrated: every night the judge’s verdict is compared against a frozen reference distribution, so a 75 means the same on any day even when the language model changes version. If we detect a shift in the model’s behavior, calibration restarts. The reasoning is the one the judge wrote before calibration.
Glossary
- Gravity (0-100)
- how much a story changes the game. 90+ is a historic event; 75-89, a structural industry shift.
- P0 · Black Swan
- an event that happens a couple of times a month globally. It interrupts.
- P1 · Tectonic
- structural shift: major regulation, $1B+ deal, frontier model.
- Branch
- a topic vertical (Frontier AI, Geopolitics, Investing…). Stories can cross several.
- Sweep
- the engine’s nightly run: ingestion, filter, clustering and scoring.
Physical-world sensors
Beyond press and research, the radar watches structured data from official public and market sources. Anomalies are detected with thresholds calibrated to each series’ historical volatility and simple statistics — no AI — and shown as hypotheses with explicit confidence, never as P0/P1.
Each source has its own cadence: markets are daily and official travel advisories are discrete events (a change of level or areas). We only show derived data from sources whose licence allows it.
- · GSCPI · NY Fed
- · OFAC · U.S. Treasury
- · CISA KEV · exploited vulnerabilities
- · U.S. Department of State · Consular Affairs
- · FCDO · GOV.UK (OGL v3)
Stories are clustered and ranked automatically. The weekly edition includes human review. Subscribe to the Weekly Brief →