FM IWA-01 · Field notes · Snapshot 04 NOV 2026
About the archive
A permanent record of what news outlets and social channels published about the 2026 Iran war, captured once, as it appeared, and kept. The archive does not explain the war. It keeps the accounts so that people can.
1Purpose
The war began on 28 February 2026. Every party to it tells a different story. Wire services report, broadcasters frame for their audiences, and state media frame for their governments. What actually happened, who said what and when, what was left out and what was quietly revised, lives in the gaps between those accounts.
Outlets edit stories. Pages go down. Archives get reorganised. This archive exists so that the accounts do not disappear: every article, kept as it was published, linked to a permanent public copy.
The machine's job is preservation. Analysis comes later, and from people.
2What is collected
23 news outlets across the parties to the war and the wider world: wire services, Western, Israeli, Arab, Turkish, Russian, Chinese and Iranian media. And 3 social sources: Telegram channels, Bluesky accounts and Truth Social.
International outlets publish far more than this war, so their articles pass a relevance gate first: a weighted keyword score covering Iran, Israel, Lebanon, Gaza, Hormuz, energy and regional impact. What the gate rejects is logged with the reason, not silently dropped.
Four outlets are kept in full, whatever the topic: Mehr News Agency, Press TV, The Cradle, Tasnim News Agency. Their whole editorial output during the war is part of the record, and when two of them run the same text, both copies are kept.
- TGMTelegram
- BSKBluesky
- TRSTruth Social
- AAMAnadolu Agency — Middle East
- AJEAl Jazeera
- MHRMehr News Agency
- JPOThe Jerusalem Post
- PTVPress TV
- TTMTehran Times
- CRDThe Cradle
- TSNTasnim News Agency
- TOIThe Times of Israel
- CNNCNN
- FOXFox News
- YNTYnetnews
- RTNRT (Russia Today)
- BBCBBC News
- APWAssociated Press
- GDNThe Guardian — Middle East
- SPKSputnik
- ABCABC News Australia
- SKYSky News Australia
- IHYIsrael Hayom
- JPMThe Jerusalem Post — Middle East
- GLTGlobal Times
- BBMBBC News — Middle East
3How it is kept
An article is captured once, at or shortly after publication. There is no re-fetch and no edit tracking: what was captured is what is kept, and the Wayback copy is the permanent citation.
- 01 Collect 198,995 items Feeds, sitemaps, channels and accounts are polled on a schedule. Full article text is extracted, not just the summary.
- 02 Preserve WAYBACK per URL Every article URL is submitted to the Wayback Machine through a persistent queue that retries with backoff.
- 03 Tag LOCAL model A locally run language model files each article under topic threads. It labels; it does not judge.
- 04 Publish DAILY snapshot Once a day a snapshot of counts, traces and bulletins is baked into these static pages. No database is queried live.
Timestamps are stored in UTC. A day in this archive runs from midnight to midnight UTC, and Day 1 is 28 February 2026. Text is normalised to UTF-8 at ingest, including the legacy encodings some outlets still serve.
4What it will not do
- No scores. Outlets are not rated, ranked or labelled for bias. The archive keeps; readers judge.
- No cross-outlet merging of state media. When two Iranian state outlets publish the same statement word for word, both are kept. Coordination is itself a fact.
- No silent drops. When a feed fails, a page will not parse or a relevance check rejects an article, the reason is logged. Gaps must be visible.
- No corrections. A captured article is not updated when the outlet edits it later.
- No tracking. This site sets no cookies, loads nothing from third parties and counts no visitors.
5Reading the instrument
The traces across this site are drawn like a seismograph's, and they measure one thing: how much was published, in 10-minute intervals, by time of publication. They do not measure how intense the fighting was. A spike means outlets were writing, not that more happened.
PRS is articles from news outlets. SOC is social posts. EVT is reserved for structured conflict-event data, which the archive does not yet collect, so it is shown as offline rather than hidden.
Heights are square-root compressed against each channel's busiest 10-minute intervals across the whole war, with one scale per channel, so a quiet day looks quiet wherever it is drawn. 4 undated articles are counted in the totals but cannot be plotted.
6Known gaps
Live collection began in week six. The first five weeks were recovered afterwards from the Wayback Machine, outlet sitemaps and other indexes, and those articles are marked as backfilled.
Feeds break. At this snapshot (04 NOV 2026, 17:00Z), 5 stations have not produced an article in over 24 hours:
- PTVPress TVlast 02 NOV 2026
- CRDThe Cradlelast 23 OCT 2026
- TOIThe Times of Israellast 02 NOV 2026
- APWAssociated Presslast 25 AUG 2026
- SPKSputniklast 20 JUL 2026
A silent station can mean the outlet stopped writing about the war, or that its feed changed and collection failed. Both are recorded; neither is papered over.
7Bulletins
Since Day 40 a language model has compiled two editions a day from the archive: a long situation report and a short brief. 414 editions have been filed. They are analysis and they have a voice; the archive does not. Every claim in them links to the archived source it came from, so they can be checked against the record.