The retained maritime intelligence experiment and its dated public snapshot, now connected to the wider EMO Atlas workspace.
Overview
The maritime pipeline and stored vessel history remain part of EMO Atlas. This page retains the maritime experiment and its existing dated public snapshot; the local Atlas dashboard has separately verified current-data connections.
The AIS Intelligence Pipeline is a local-first maritime intelligence experiment. It continuously receives AIS vessel observations and stores them as structured history rather than treating each message as an isolated event.
The motivation is that reasoning about the world requires history. A single snapshot tells you where a vessel is; a continuous record tells you where it has been, how it behaves, and what is unusual. The pipeline gives downstream reasoning something to work from.
This is an active experiment running on consumer hardware. It is helping us understand the cost of continuous ingestion, the value of accumulated state, and how local infrastructure can support higher-level intelligence without depending on a remote API for every decision.
EMO Atlas / Maritime History
EMO Atlas receives maritime observations and preserves them over time. This public view combines dated aggregate history with a separate vessel-position snapshot. It does not connect to the private collector.
Dated Aggregate Snapshot
Exported October 6, 2026 at 10:57 p.m. Toronto. Cached history ends 2026-10-04 (2026-10-04 UTC). The latest collector count of 8,251,691 stored observations is separate from this historical dataset and is newer than the cached history.
Daily received reports in UTC. Select a day to inspect volume. Zero reports can reflect a collection gap; they do not establish zero vessel traffic.
2026-10-04 UTC · 245,180 reports · 2,079 identifiers
Cached history 2026-09-04 04:20 to 2026-10-04 21:16 UTC; analysis created 2026-10-04T21:17:59.820256+00:00. Current collector totals are newer than this cached history.
Drag to pan; use + / − or pinch to zoom. Each marker is an actual AIS position received by the collector, not an aggregate cell. Select a marker for coordinates and its receipt time.
Position snapshot: 2026-10-06, 11:25:03 p.m. Toronto · 1,669 vessels · 0 min since export. Gold: received within 15 minutes before export; grey: 15–60 minutes. This is a static snapshot, not current ship locations.
Basemap: OpenStreetMap. Positions: AISStream; timestamps are collector receipt times. The historical day slider above does not change this separate position snapshot. No identities or private API endpoints are published.
| Last receipt · UTC | Next receipt · UTC | Gap |
|---|---|---|
| 2026-09-11 12:21 | 2026-09-18 02:17 | 157.93 h |
| 2026-09-05 16:02 | 2026-09-06 05:20 | 13.29 h |
| 2026-09-04 18:52 | 2026-09-05 02:24 | 7.53 h |
| 2026-09-28 03:03 | 2026-09-28 03:21 | 0.30 h |
| 2026-09-25 02:23 | 2026-09-25 02:40 | 0.29 h |
| 2026-09-28 05:12 | 2026-09-28 05:29 | 0.28 h |
| 2026-09-26 04:54 | 2026-09-26 05:11 | 0.28 h |
| 2026-09-05 08:54 | 2026-09-05 09:11 | 0.28 h |
AIS observations support movement history, but cannot establish cargo, motive or a complete census. Receipt timestamps can differ from transmission time. The map shows individual received positions from a separate snapshot. Historical individual-vessel playback is not available in this public edition. The local dashboard keeps the private live workflow.
EMO Atlas / Caribbean energy evidence prototype
Selected local observations alongside dated primary-source reporting. This public prototype does not expose the local assistant or detailed analytics.
Static snapshot. Data as of 2026-10-08, 10:28:50 p.m. Toronto · 0 min old. Exported 2026-10-08, 10:41:19 p.m. Toronto. Automatic hosted updates are not connected.
Stored observations, rounded
Last-hour vessels, rounded
Derived voyage records
Classified tanker voyage records
Seven derived voyages are stored; zero are classified as tanker voyages in this small record set.
Week of 2026-09-28 · at least 72 hours delayed · no identities or precise locations
One delayed record links two geofenced locations within Trinidad and Tobago. Vessel identity, exact ports, coordinates and timing are omitted. A recorded movement supports testing the workflow; it does not establish an oil cargo or international oil flow.
Reported projection · Published 2026-05-06 · Event/update 2026-05-05 · 2026/2027 financial-year projection
JIS reports a 12.21 million-barrel sales projection, including domestic and export allocations. This is a plan, not independently measured shipments.
Reported shipping conditions · Published not stated on rolling page · Event/update 2026-10-08 · Incident list updated October 8; selected incidents include October 5
IMO publishes incident and shipping updates for the Strait of Hormuz. The reported disruption supplies global context; it does not prove a change in this Caribbean sample.
Official statistical release · Published 2026-09-30 · Event/update not an incident · July 2026 data
The release covers July petroleum supply and disposition. Publication lag and different geographic coverage prevent a same-day comparison with local AIS.
Stored position range: 2026-09-04 to 2026-10-08. Caribbean receiver coverage; gaps and uneven reception not quantified. AIS does not prove cargo, sanctions violations, causality or forecasts. Counts are not a census; freshness and source coverage differ. No proprietary M4 analytics are exported.
The local tracker now has a read-only Hermes companion. Typed commands refresh evidence, search saved teacher notes, open the local view and save drafts, with separate queued/completed/error beeps. A browser summary completed in 2.25 seconds. A bounded, cited model explanation completed in 89 seconds without executable tools.
Five grounding and safety tests passed. Installed macOS speech playback produced a valid private test recording. Microphone recognition remains unavailable because the Whisper cache lacks model weights. Durable Markdown lessons connect verified workflows to the existing Obsidian architecture note; model weights were not retrained.
The next test separates small selected exports from incoming ship-feed traffic. A public upload adapter is still needed before a five-minute hosted cadence can be called live. The collector, detailed database and M4 computation stay local.
Related Projects