agenticinfection

Which diseases are public health authorities publishing about right now — ranked by how many independent authorities are talking, never by how bad anything is.

GENERATED 2026-08-22 09:47 UTC (our clock) · WINDOW 30 days · NEWEST ITEM 2026-07-30T11:00 (the source's own date)
If the GENERATED stamp above is old, this observatory has stalled. That is a fact about us, not about disease activity — and it is the one thing a page like this must never let you mistake for quiet.
Not medical advice. This page reports how often public health authorities are PUBLISHING about a disease. It does not report how much disease there is, how dangerous it is, or how worried anyone should be, and no ranking here should be read as a risk assessment. It is not medical advice and it is not a substitute for your own public health authority or clinician. For guidance that applies to you, go to the authority for your country — several are linked in every row below.
HaoHung Dang — MSc in Public Health (UBC, Vancouver)

First report

Diseases whose earliest item in this window is also the earliest we hold at all. A new outbreak arrives as one item from one authority — so it can never rank highly on the list below, which orders by how many authorities are publishing. This list exists because that is a property of the ordering and not a property of the disease.

e-coli

First seen 2026-08-21T08:00 · 2 source(s) so far

rsv

First seen 2026-08-13T00:01 · 1 source(s) so far

measles

First seen 2026-08-08T21:09 · 1 source(s) so far

orthoebolavirus/bundibugyo

First seen 2026-08-01T10:13 · 2 source(s) so far

hepatitis-b

First seen 2026-07-29T13:14 · 1 source(s) so far

hiv

First seen 2026-07-27T19:08 · 1 source(s) so far

salmonella

First seen 2026-07-27T12:07 · 3 source(s) so far

cyclospora

First seen 2026-07-24T21:00 · 2 source(s) so far

The ranking

1 cyclospora

2 independent bodies published about this in the window · 2 items · newest 2026-07-30T11:00

Resolved from these source strings: cyclospora ×2 — shown so a wrong grouping is visible to you rather than only to us.

2 salmonella

1 independent body published about this in the window · 9 items · newest 2026-08-21T15:35

How this count was reached: us-cdc published through 3 of its own channels (cdc_mmwr, cdc_newsroom, cdc_outbreaks), counted once. A body speaking through several of its own channels is one body. We show the collapse so you can check it.

Resolved from these source strings: salmonella ×9 — shown so a wrong grouping is visible to you rather than only to us.

3 e-coli

1 independent body published about this in the window · 2 items · newest 2026-08-21T15:35

How this count was reached: us-cdc published through 2 of its own channels (cdc_newsroom, cdc_outbreaks), counted once. A body speaking through several of its own channels is one body. We show the collapse so you can check it.

Resolved from these source strings: e. coli ×2 — shown so a wrong grouping is visible to you rather than only to us.

4 orthoebolavirus/bundibugyo

1 independent body published about this in the window · 4 items · newest 2026-08-18T10:33

How this count was reached: who_afro relayed who rather than reporting independently. A body speaking through several of its own channels is one body. We show the collapse so you can check it.

Resolved from these source strings: ebola disease caused by bundibugyo virus ×2, ebola bundibugyo virus disease ×2 — shown so a wrong grouping is visible to you rather than only to us.

5 rsv

1 independent body published about this in the window · 1 items · newest 2026-08-13T00:01

Resolved from these source strings: rsv ×1 — shown so a wrong grouping is visible to you rather than only to us.

6 measles

1 independent body published about this in the window · 1 items · newest 2026-08-08T21:09

How this count was reached: paho_news relayed who rather than reporting independently. A body speaking through several of its own channels is one body. We show the collapse so you can check it.

Resolved from these source strings: measles ×1 — shown so a wrong grouping is visible to you rather than only to us.

7 hepatitis-b

1 independent body published about this in the window · 1 items · newest 2026-07-29T13:14

How this count was reached: paho_news relayed who rather than reporting independently. A body speaking through several of its own channels is one body. We show the collapse so you can check it.

Resolved from these source strings: hepatitis b ×1 — shown so a wrong grouping is visible to you rather than only to us.

8 hiv

1 independent body published about this in the window · 4 items · newest 2026-07-27T21:58

How this count was reached: paho_news relayed who rather than reporting independently. A body speaking through several of its own channels is one body. We show the collapse so you can check it.

Resolved from these source strings: aids ×2, hiv ×2 — shown so a wrong grouping is visible to you rather than only to us.

What this page knows about itself

Population217 distinct items from 11 sources
Qualifying8 entities in this window. A quiet window says so.
Unresolved12% extracted names we could not map to one disease. Retained, not discarded — the list is how the vocabulary grows.
No candidate66% titles containing no name we hold. These never reach the resolver, so the unresolved rate cannot see them.
Ambiguous0 names claimed by more than one disease (hepatitis). Counted toward none of them, on purpose.

These five numbers are published for the same reason the ranking is: a ranking that does not report its own coverage presents a partial population as a complete one. If they are bad, they are visible.

Coverage — what is in and what is not

Generated from the source registry on every build. It is not a promise about scope; it is a list of what the pipeline actually reads today.

AuthorityRegionTierStateNote
WHOglobalT-NEWSINTHE AUTHORITATIVE T-NEWS SOURCE, and it needs NO browser runtime. Public OData endpoint behind the DON page. FOUND BY MEASUREMENT after the page itself yielded zero item hrefs. $or
WHOglobalT-NEWSNOT YETREFERENCE CASE. 330KB of HTML and ZERO item hrefs -- the chrome renders server-side, the DON LIST does not. My classifier calls it HTML because nav boilerplate clears the visible-t
WHOglobalT-NEWSNOT YEThistorical WHO DON feed path; UNVERIFIED
WHOglobalT-NEWSIN
WHO AFROafricaT-NEWSIN
PAHOamericasT-NEWSINCORRECTED Day-166: /en/rss/news was a guess and 404'd; /en/rss.xml returns 10 entries.
US CDCusT-NEWSINCDC Online Newsroom. CORRECTED Day-166: my first candidate (media/404952) is a live, well-formed, ZERO-ITEM COVID topic card -- a FEED-SHAPED NULL, kept below as a named reference
US CDCusT-NEWSIN
US CDCusT-NEWSIN
ECDCeuT-NEWSNOT YETBOTH ECDC guesses 404. Its Drupal rss.xml IS taxonomy-scoped -- /en/taxonomy/term/1490/feed returns 10 entries of PROCUREMENT NOTICES, so the mechanism works and the TERM ID is wha
ECDCeuT-NEWSNOT YET404. See ecdc_news.
UKHSAukT-NEWSIN
PHACcanadaT-NEWSIN
AU HealthaustraliaT-NEWSNOT YETRead timeout on three separate attempts from this host. UNMEASURED, not absent (2b.5).
Africa CDCafricaT-NEWSNOT YET
ProMEDglobalT-NEWSNOT YETNo feed at /feed/, /promed-posts/feed/, or wp-json. ProMED is subscription-gated; this is a P2.2 LICENSING question, not an engineering one, and it must be answered before any rout
CIDRAPglobalT-NEWSIN
ONTglobalT-NEWSIN
EurosurveillanceeuT-NEWSNOT YET/rss/ returns a well-formed feed with ZERO items -- FEED_EMPTY, a live second instance.
gov.ukukT-NEWSNOT YETplaceholder query; probe will show whether a bare query returns a feed
NCBIglobalT-EVIDENCENOT YETE-utilities takes BOOLEAN queries natively -- anh's systematic-review method as an API. BUT robots.txt DISALLOWS /entrez/eutils/ for our UA. NCBI's API terms permit programmatic us
EuropePMCglobalT-EVIDENCENOT YET
medRxivglobalT-EVIDENCENOT YET
US CDCusT-NEWSNOT YETREFERENCE CASE, not a working source: 200 OK, valid RSS 2.0, lastBuildDate 2026-06-24, ZERO items. Retained so the FEED_EMPTY verdict has a live instance and cannot decay into folk
US CDCglobalT-POLICYNOT YETTRAVEL HEALTH NOTICES. The strongest T-POLICY row found:每 entry is an ACTION by a named authority, dated, naming a disease and a country ('Level 2 - Zika in Indonesia'). ⚠ THE LEVE
UK FCDOglobalT-POLICYNOT YETUK travel advice, per country. Same class as cdc_travel_notices from a second government -- two independent advisory streams is what makes a policy-velocity signal mean anything.
US FDAusT-POLICYNOT YETAPPROVALS AND AUTHORISATIONS. anh's countermeasure/market-access layer starts here: an approval is an EVENT with money attached (MucM 5.3).
US FDAusT-POLICYNOT YET
Canada GazettecanadaT-POLICYNOT YETRegistered regulations. HIGH VOLUME AND MOSTLY IRRELEVANT -- carried because a regulation is the hardest form of policy action and the noise is the price. Its no-candidate rate wil
ECDCeuT-POLICYNOT YETFound while hunting ECDC news. Procurement is a T-POLICY class anh ruled in day one and AMR economics is the standing beat.
EMAeuT-POLICYNOT YET404 Day-166. EMA is the European half of the approvals layer and its absence is a REAL GAP, named rather than quietly dropped. Route unknown; not guessed further.
CFIAcanadaT-POLICYNOT YET⛔ 404 Day-166 -- AND THIS URL IS LANDED IN THIS REPO. 2-FLEET/data/kits/canadian-health-feeds.yaml dispatches a crawl to it as an 'authoritative' source. A KIT THAT POINTS AT A DEA

Method

Rank key. Primary: the number of distinct sources that published about an entity inside the window. Tie-break: the newest item's date. Item count is shown and is deliberately not ranked on — one authority publishing eight updates is one authority talking; three authorities publishing once each is three authorities agreeing.

What the rank never reads. Popularity, engagement, shares, our own click data, editorial judgement of importance, or any severity proxy — deaths, case counts, the words "serious" or "urgent". This is structural rather than promised: nothing in the input carries severity, so nothing in the output can.

How a name becomes an entity. Titles are scanned for occurrences of known disease names at word boundaries. That is a lookup, not a guess: a name we do not hold is never inferred. The consequence is that we miss rather than invent, and the miss rate is the "no candidate" number above.

Why there are two lists. A disease that has just been reported for the first time has, by definition, one item from one authority. On a list ordered by how many authorities are publishing it must sit at the bottom — on the day it matters most. That is a property of the ordering, not of the disease, and no re-weighting fixes it without smuggling a judgement about importance into a measure of signal. So the first-report list above answers a different question and is ordered by when we first saw it.

What the rank key can and cannot separate. Measured on the live corpus: six entities are carried by more than one authority — the top of this list really does mean several independent bodies are publishing about this. Seventeen are carried by one. Those seventeen all tie on the primary key, so below the top of the list the tie-break — recency — is doing the ordering. The distinct-source count on every row tells you which regime that row is in: a row showing 1 is ordered by how recently it was published, not by how many authorities agree. We are saying so here because the ranking looks identical in both regimes and only this sentence separates them.

Dates. Each row carries the source's own publication string and the time we retrieved it. An item whose date cannot be parsed is excluded from the window and counted — never dated to now. We do not publish a page-level "updated today": a freshness claim on the page tells you nothing about the freshness of the rows.

Excluded. Non-T-NEWS rows; items outside the window; retrievals of an item we already hold (we count items, not our own polling).

Corrections

If a row is wrong — a mis-grouped disease, a stale link, a source we are reading badly — say so and it gets fixed in public. ⛔ NO CORRECTION ROUTE IS PUBLISHED YET. We are not offering one we cannot answer: an offer implies a reader. When the route exists it will be the same address as the privacy contact, because two addresses would mean one of them is decoration.

Every entity above lists the exact source strings that were grouped into it. That is there so a wrong grouping is legible to a reader, not only to us: the unresolved count measures what we missed, and the grouping list is the only thing that can show what we joined wrongly.