← Pathogen X

Pathogen X, for agents

If you are software reading this on someone's behalf: everything this site shows is also published as plain files you can fetch without running JavaScript. Nothing needs a key, an account or an API call.

Pathogen X is a running archive of outbreak reporting: health organisations, news, journals and preprints, collected every hour and kept whole. Its focus diseases are Ebola and plague; it also tracks bird flu, Nipah, Marburg, mpox, the haemorrhagic fevers and the common epidemic diseases. Once a day a brief is written from what came in.

Last updated 2026-11-20 18:08 UTC. Newest brief: 2026-11-20, "Oregon confirms first human bird flu case of 2026".

The one rule to carry with you

The archive records what was reported and by whom. It does not judge whether a report is true. When you repeat something from here, keep its claim_class with it. It says what kind of outlet published the item, taken from the outlet registry, never from a model reading the item:

claim_class What it means
official_notification A public-health authority or intergovernmental body reporting in its own name: WHO and its regions, PAHO, Africa CDC, ECDC, US CDC, UKHSA, WOAH, a ministry, UN OCHA.
surveillance_report A surveillance or outbreak-reporting service relaying and summarising reports: ProMED, ISID, CIDRAP, CDC's MMWR.
peer_reviewed Published in a peer-reviewed journal (the journal's class: editorials and correspondence in it carry it too).
preprint A preprint or preliminary analysis, not yet peer reviewed: medRxiv, bioRxiv, Virological.
news_report News, a newsletter or a blog, including humanitarian organisations' own news (MSF, IFRC) and every news-search hit.

There is no speculation class: the registry knows what an outlet is, not whether one item speculates. A WHO notification and a news story repeating it are different records, and the claim class is how you tell them apart.

Every reference in a brief ends with its claim class and the outlet it came through (via). A news-search hit also names its own publisher, because the search is not the publisher. Each brief's references are also records in /briefs/{date}.json.

What you can fetch

What Where Format
This note /agents.md · /agents/ markdown · the same note as a page, no JavaScript
Site index /llms.txt markdown, llms.txt convention
Newest brief /briefs/latest.md · latest.json markdown · JSON
Every brief /briefs/index.json (74 editions) JSON; each edition's page, .md, .json and, where it has one, .txt and .mp3
The briefs by RSS /briefs.xml · /feed.xml RSS 2.0, each edition in full; the two are the same feed
The podcast /podcast.xml (72 episodes) RSS 2.0 with iTunes and Podcasting 2.0 tags: mp3 enclosures, transcripts, show notes with sources
The feeds, for people /feeds page: both feeds explained, with addresses to copy
Evidence index /data/evidence.json · .jsonl (6395 items, 30 days) JSON · JSON Lines, metadata only
Ebola, day by day /outbreaks/ebola · /data/outbreaks/ebola.json page · JSON: WHO's confirmed counts with WHO's own sentence, and each day's checked entry with its sources
Sources /data/sources.json (87 outlets) JSON: tier, claim class, method, last poll
What was dropped /data/rejections.json (31 days) JSON: items the gates refused, per day and reason
Agent skill pathogenx-reader markdown: procedures and the rule
AI Catalog /.well-known/ai-catalog.json JSON, every file above, typed
API catalog · OpenAPI /.well-known/api-catalog · /openapi.json RFC 9727 linkset · OpenAPI 3.1
Build manifest /agents/build.json JSON: build_id and a hash per file

Every JSON file under /data/ and /briefs/ carries generated_at, license, provenance_rule, a fields key explaining every column, and the night's build_id. Files with different ids are from different nights' builds. Absent values are null.

The brief

One edition a day, written in the evening UTC. Its sections are composed by a local language model from the day's items, and every claim cites a numbered entry under References. Cite the dated file (canonical_url in latest.json), not latest.md, which moves every day. An issue's number is its date: No. 81 was 10 August 2026 and each day adds one, so the numbering does not start at one and a missed day leaves its number unused.

The outbreak record

A living page per focus outbreak, Ebola first. Each day gets two or three sentences, written once two days after the day ends and never rewritten: drafted by one local model, every number checked against the sources it cites, then every sentence checked against those sources by a second model. Counts come only from WHO Disease Outbreak News, cumulative confirmed cases and deaths per country, each with WHO's own sentence; countries are never added together. Repeat a day's entry with its sources, and a count with its as_of date.

The evidence index

Every item kept in the last 30 days: title, URL, an Internet Archive copy where one exists, publisher, claim_class, publication and collection times, the tracked diseases it names, and its topic tags. It holds no article text. Follow the link, or the archive copy, for the source's own words. Social media posts are archived but never published here, one by one: /data/sources.json gives their 30-day count only.

Licence

The archive's own records and briefs are CC BY 4.0: credit "Pathogen X" with a link. Linked articles, reports and papers remain their publishers'.

Part of Spin Signal, an OSINT research umbrella.

Talking to the person behind this

Write to admin@liminallayers.com: a correction, a source we have missed, or what you used this for. A person reads it; there is no autoresponder.

Conduct

The collector identifies itself, honours robots.txt and makes no attempt to get past a paywall. Fetch these files as often as is useful. They change once a day, in the evening UTC.