Open-source intelligence

Nothing here is broken into. It is already published, scattered across providers and records that were never meant to be read together. Ossintheus puts them in one graph and keeps the receipt for every line.

What's on board

01 — Sweep

Every provider at once

One identifier goes to the whole catalogue in parallel, not a sample of it, and every hit is re-run against usernames that cannot exist before you are shown it.

02 — The core

See the world unassembled

03 — The map

The map

Cameras, aircraft and other located signals as composable layers, each carrying its own source, its own time, and its own attribution.

  • 126,000

    Public cameras in the hosted catalogue

  • Live

    Aircraft positions, refreshed on a cache

  • Layers

    Composable, each with its own attribution

What the map does

On the map

Movement, imagery and hazard, each on its own layer with its own source and its own clock — so every one of them answers where it came from and when it was true.

  1. Live vehicle tracking

    Camera to camera as the hits land. Every track carries its own confidence — the working rate is 75%, and anything under the bar stays a candidate.

  2. Live plate search

    One plate against every camera in scope, back as the frames it was read in, each stamped camera, place, time.

  3. Keyword search across cameras

    Name a colour, a vehicle, a scene. The model watches the cameras you choose and returns what it matched.

  4. Image geolocation

    No metadata needed. Terrain, skyline, signage and shadow are read together — to the metre where the frame allows.

  5. It says when it cannot

    A frame too thin to place returns the reason it is thin, never a guess wearing a coordinate.

  6. Your own data as a hint

    Feed in what the case knows. It narrows the search, stays labelled a hint, and is never promoted to evidence.

  7. 150,000+ vessels

    Cargo ships, tankers, fishing boats and jetskis on live feeds, each keeping the track behind it.

  8. Aircraft, in real time

    Every aircraft on the feed, updated as it moves, with the age of the last position on the layer.

  9. Weather and conditions

    Conditions and live weather events, beside the cameras and the aircraft they fell on.

  10. Wildfires and conflict zones

    Fire detections with their perimeters, conflict areas from published reporting. Both dated, because both move.

  11. World resource monitor

    Energy, mining, shipping lanes and the infrastructure under them, as layers beneath everything else.

What is on board

The armory

A package per technique, a file per site — which is why the catalogue keeps growing without the product changing shape. Open a module for what it carries; every count on this page has a receipt.

  • One sweep takes a handle to every provider in the catalogue at once, then fuses the hits into a single identity panel with per-value site counts.

    • 3,434 providers per sweep
    • One fused identity panel
    • Per-value site counts

  • Registration checks across the service catalogue, plus direct probes against mail providers. Unconfigured probes refuse to run rather than guess.

    • 388 sources per address
    • 109 mail domains probed directly
    • Unconfigured probes refuse

  • Name history, guild sightings, verification records and the enforcement trail, streamed as they resolve — a slow module is not a broken one.

    • Name history
    • Guild sightings
    • Enforcement trail

  • Lookup across a held index of more than 100 million records: handles, numeric identifiers, group and channel sightings, and the history behind a name. Every hit names the record it came out of.

    • 100M+ records indexed
    • Handle, ID and group sightings
    • Every hit names its record

  • Carrier and line facts are held apart from account-existence checks, so a technical property is never read as evidence about a person.

    • 115 sources per number
    • Carrier facts kept separate
    • Line type, not conclusions

  • DNS, WHOIS, TLS, HTTP and file metadata collected by separate collectors, each recording what it observed rather than what it concluded.

    • DNS · WHOIS · TLS
    • HTTP · file metadata
    • Observations, not conclusions

  • Templates rendered for review in a shallow run, executed in a full one. You can always read the query before it is sent.

    • 83 templates
    • Rendered before run
    • Every query readable

  • Providers measured rather than assumed: availability and coverage are checked, not taken from documentation.

    • Availability checked live
    • Coverage measured
    • Nothing taken on faith

  • Cameras, aircraft, vessels and other located signals as composable layers, each carrying its own source and attribution. Vehicle and plate tracking, keyword search across the cameras you choose, and metadata-free image geolocation run on top of it. Everything drawn is also a searchable row, so the surface works by keyboard.

    • 126,000 cameras hosted
    • 150,000+ vessels · live aircraft
    • Geolocation, weather, hazards

    Everything the map does

Where these numbers come from Every figure on this page, and what it was counted from. 8 figures
3,434 providers
Counted: the length of the services array in services.json, the catalogue a username sweep iterates. 16 August 2026.
388 email sources
Maintainer figure. The public repository exposes only the mail-provider map (109 domains across 62 providers); the full source list is not separately countable from it.
115 phone sources
Counted: six hand-written sources in modules/phone/sweep.py plus 109 rows in data/phone_endpoints.json.
83 dork templates
Counted: the length of templates in dorks.json.
126,000 cameras
Maintainer figure for the hosted catalogue. The open repository ships a smaller set of roughly thirty thousand.
100M+ Telegram records
Maintainer figure for the held index. It is not countable from the public repository, which ships no part of that index.
150,000+ vessels
Maintainer figure, counted as the distinct vessels seen on the position feeds the map subscribes to. It moves with the feeds.
75% vehicle-track confidence
Maintainer figure: the working rate the tracker is held to before a track is drawn as a track. Every track also carries its own score on the map, and the score, not this figure, is what a reader should go by.

Counted figures were taken from the application's own data files on the date shown. Maintainer figures are stated by the people who run the service and are not verifiable from the public repository — they are labelled that way rather than presented as measurements. All of them drift as sources are added and removed, and if a figure here disagrees with the application, the application is right.

The workspace

Open the Instrument

Open the workspace dash.ossintheus.app
Build
The one described above — not a demo of it
Access
Sign-in required
Sharing
Expiring read-only case links — no account needed to read one

Tour A walkthrough of the workspace goes here.