AI Analytics

An independent research and engineering studio. We build measurement infrastructure for the open internet, primary-source federal reference systems, and post-quantum communications — led by curiosity and shipped as open data or open-source. Operating since 2020.

Three flagship projects span different domains: an internet-censorship observatory across 200 countries (Voidly), a unified reference for US federal regulatory data (Federal Data Hub), and a post-quantum encryption SDK for autonomous drone fleets (Swarm). Beyond the flagships we run an OSINT pipeline at scale, statistical election-anomaly tooling, and a steady stream of technical writing and intelligence briefs. Work is self-directed and cross-domain by design.

Across our projects

Countries observed
200
Voidly measurements
2.2B+
Verified incidents
1,574+
Probe nodes
37+
Federal datasets
144
Canonical records
12.3M+
Swarm tests
465
Years operating
6

Voidly stats from the live measurement network; federal-data stats refreshed at build time from api.ai-analytics.org/coverage.

Projects

Voidly — Global Censorship IndexOpen data · CC BY 4.0
Real-time, evidence-backed measurement of internet censorship across 200 countries. ML-based detection of throttling, DPI filtering, and routed BGP blocks; 7-day shutdown forecasting; a public dataset purpose-built for journalism, research, and human-rights work. 2.2 billion network measurements, 1,574 verified incidents catalogued to date. Probes across 37+ autonomous-system vantage points.
US Federal Regulatory Data HubOpen data · CC0 1.0
Unified reference for US federal regulatory records — SEC, FDA, OFAC, DOJ, EPA, CFPB, IRS, USAspending, NHTSA, FAA, CFTC, FEMA, NOAA, CDC, NIST, and 30+ other agencies. 144 datasets, 12.3M+ canonical record URLs, daily refresh, REST + JSON-LD + Model Context Protocol surfaces. One query returns every regulatory event for a company across all datasets — example: ExxonMobil cross-agency timeline.
Swarm SDK — Post-Quantum Drone EncryptionProprietary
Python SDK for encrypted communications between drones and counter-UAS systems. ML-KEM-768 + X25519 hybrid key exchange, MAVLink v2 transport, gossip-mesh routing, forward secrecy and key rotation under contested-RF conditions. CNSA 2.0 compliant. 465 tests across seven encryption modes; Python 3.9 through 3.13.

Adjacent work

Internal tooling and infrastructure that supports the public projects. Not separately released as products; documented here for completeness.

Footprint Vault
OSINT reconnaissance across 100+ public platforms; offline privacy-audit mode and enterprise threat-intel mode.
OSINT pipeline
Real-time social-media ingestion at 2.4M posts per hour. Kafka, TimescaleDB, NLP scoring across 80 GPU workers.
Election monitoring
Statistical anomaly detection across voter, turnout, and campaign-finance data — Benford's Law, ARIMA, ML turnout models.

How we work

AI Analytics is run by a small group of independent operators across time zones. Projects are picked by interest, not roadmap. Output biases toward open data, open source, and primary-source citation. We respond to press, research, and security inquiries within 24–48 hours.

More background: about · methodology · governance and editorial policy · writing · intelligence briefs · six-year timeline · contact (PGP, Tor).