Is this site written by AI?

The selections and the structure are produced by software; the definitions of the entities are written by hand. Each briefing is assembled by a build process that reads a monitored feed, groups items by the organisations and products they mention, and orders them by a published scoring rule. No language model rewrites the items — they are reproduced as recorded, or not shown at all. The one place a person writes prose is the entity definitions, which are marked as such and are absent for any entity the desk cannot describe from knowledge it trusts.

Why do items have no link back to the original post?

Because the feed supplies one that does not resolve. In the window this site was built from, every one of the 1209 items carried the same placeholder URL rather than a permalink to the post it came from. Rather than link every item to a page that is not its source, briefings carry no per-item source link at all. Where an item contains a real URL in its text, that URL is shown and marked nofollow.

Why is there no journalist’s name on the briefings?

Because there is no journalist. The briefings are compiled by the editorial desk from the recorded feed, and the author is given as the desk rather than as a person. The feed’s own author field is not used: in the current window every item carried the same author string regardless of who actually wrote it, so publishing it would attribute many people’s words to one name. Adding a named editor with real credentials would strengthen this site; inventing one would not.

Do you verify what you publish?

No, and the briefings say so on every page. An item is recorded as the feed captured it and is not independently checked against the company’s own announcement. What the site does verify is what it can: entity identities, official channels, and the arithmetic on each page. Claims about products and releases should be confirmed at the official channel linked from the relevant entity page.

How do I get something corrected?

Write to editor@bloger.fm with the briefing date and the item. Corrections that the feed itself got wrong cannot be fixed at the source, so the item is removed from the briefing when a correction is confirmed, and the briefing's modification date changes. Definitions and entity metadata are corrected directly and take effect on the next build.

What does “gaining ground” mean?

It means an entity was mentioned more often in the second half of the window than the feed as a whole grew — so it took a larger share of coverage than it had. It is not a claim that the company is doing well. The comparison is against the feed's own growth because the feed roughly doubled across this window: measured against raw counts, almost everything looks like it is rising, which measures the feed and not the entity. Full method, including the thresholds, is in "What the site works out, and how" above.

Editorial policy

How this briefing is made

Bloger.fm is an aggregator, and this page describes exactly what it aggregates and what it does with it — including the parts of the source data that are unreliable, and what the site does instead of pretending otherwise.

Applies to data synced 2026-09-20 · window 7 days · 1200 items

What this site publishes

Two things, and nothing else. Daily briefings, one per calendar day, which file everything recorded that day under the organisation or product it concerns. And entity pages, one per organisation, product or standard, collecting everything recorded about it across the whole window.

There are no opinion pieces, no reviews, no sponsored posts and no affiliate links. The site has no advertising and no commercial relationship with any company it covers.

Only complete days are published. A briefing is a day's news, so the day currently in progress is held back until it finishes — the newest briefing here is always yesterday or older. Publishing a day at midday would mean a page with the morning on it, indexed as the day's record and then rewritten underneath whatever a search engine had cached.

How a briefing is assembled

01 Sync. A build reads the last 7 days from an upstream PostgreSQL feed of public AI announcements and writes it to a snapshot. Nothing downstream of this step talks to the database.
02 Tidy. Each item’s text is cleaned of pipeline artefacts only — link syntax, dead URL shorteners, doubled whitespace. Nothing is paraphrased, reordered or trimmed. A quote-tweet is split so the quoted post is shown as a quote and never as the poster’s own words.
03 Match. Items are matched against a hand-maintained dictionary of 49 entities. Matching is on word boundaries and longest-name-first, so “Claude Code” is not also counted as “Claude”. An item that names one or two entities is treated as being about them; one that names six or more is treated as a roundup and ranked down.
04 Order, then publish. Within a section, items are ranked by a published formula: a baseline, plus substance up to a cap, plus a bonus for naming one or two entities, plus a bonus for announcing a change, minus a penalty for being a roundup. The formula decides reading order only — it never includes or excludes an item.

What the site works out, and how

Briefings and entity pages carry a handful of judgements that are not in the source feed — which entities are gaining or losing ground, which day was busiest, what is being recorded for the first time. All of them are computed, and all of them are stated below so a reader can disagree with the method rather than only with the conclusion.

“Gaining ground” / “losing ground”: The window is split in half and each entity's item count in the two halves is compared — but against how much the whole feed grew, not against the entity's own earlier count. This matters more than it sounds. In the window this page describes, the feed more than doubled between its halves; comparing raw counts made 26 entities look like they were rising against 1 falling, which was a measurement of the feed's growth and not of anything the entities did. An entity is now rising only if it outgrew the feed, and falling only if it lost ground while the feed gained. A change also has to clear an absolute floor of five items, so a move from two to three does not register as a trend. An entity whose first appearance falls after the window opened is reported separately rather than as a trend, because there is nothing to compare it against.
“New to this window”: The entity's first appearance falls in the second half of the window — it is absent from every briefing before the one it is marked on, and from the entire first half. The stricter rule is deliberate. An earlier version marked anything absent from the window's opening day, which flagged ten entities as new, seven of them for merely missing the feed's quiet early days: this feed carries roughly a third of its later volume in its first four days, so an entity that simply arrived after the ramp looked like a debut. One entity passed the stricter test in the window this page describes.
What “new” is not: It is a statement about this site's coverage, not about the organisation or product. The window is 7 days wide, nothing older is retained, and an entity marked this way may have been covered elsewhere for years. The label reads “new to this window” rather than “new” for that reason, and the entity's page states the exact date its coverage in this window begins.
“Busiest day”: Of the days in this window, this one has the most items mentioning that entity. Only shown for entities recorded on more than one day — with a single day there is nothing to compare, and the marker would appear everywhere and mean nothing.
Reading order: Sections and items are ordered by a weighted score: a baseline, plus substance up to a cap, plus a bonus for naming one or two entities, plus a bonus for announcing a change, minus a penalty for being a roundup that names many entities and is therefore about none of them in particular. The score decides reading order only. It never includes or excludes an item — every item in the window appears on its day's briefing.

One limit worth stating plainly: these are counts of mentions in a feed, not measures of importance. A company having a quiet week and a company being written about less look identical here. The numbers are reported as what they are — how much this feed said about something — and nothing on the site upgrades that into a claim about the world.

What the source data gets wrong

Three columns in the upstream feed are unreliable. Because the whole site is built from that feed, each one is handled explicitly rather than passed through.

The author column is not a byline: Every item in the current window carries the same author string — 4,984 items under one name, in a mix of languages, covering every company in the industry. It is a collector’s label, not the person who wrote the post. Publishing it would attribute thousands of people’s words to one individual, so it is not published at all. The site bylines the editorial desk instead.
The URL column is a placeholder: All 1200 items in the current window carry the same URL, pointing at a post rather than at the item itself. There is therefore no per-item source link anywhere on this site. When an item’s own text contains a real URL, that URL is shown and marked nofollow.
The summary column contradicts the text: The feed attaches a machine-written rationale to each item, and spot-checking finds it describing different news from the text it is attached to. It is kept in the snapshot for debugging and is never shown to a reader.

Two kinds of entity

36 of the 49 tracked entities have a written definition — a stable sentence about what the thing is, checked by hand, that should still be true in a year. These carry an official channel and a structured-data identity.

The other 13 are tracked but undefined: names the feed uses constantly that post-date anything the desk can verify, or that have no stable public identity at all. Their pages carry recorded coverage and say plainly that no definition has been written. They emit no structured-data identity, because there is nothing verified to point at. An entity is promoted to defined — gaining a definition, an official channel and a sameAs — once that becomes possible.

Corrections

Email editor@bloger.fm with the briefing date and the item. Because a briefing is a record of what the feed captured, a factual error in the feed cannot be edited at source — a confirmed correction removes the item and the briefing’s modification date changes. Errors in a definition or in entity metadata are fixed directly.

The site keeps no separate corrections log. With a rolling 7-day window and no archive beyond it, a corrected briefing is superseded within the week, and a log of changes to pages that no longer exist would be noise rather than accountability.

Common questions

Is this site written by AI?: The selections and the structure are produced by software; the definitions of the entities are written by hand. Each briefing is assembled by a build process that reads a monitored feed, groups items by the organisations and products they mention, and orders them by a published scoring rule. No language model rewrites the items — they are reproduced as recorded, or not shown at all. The one place a person writes prose is the entity definitions, which are marked as such and are absent for any entity the desk cannot describe from knowledge it trusts.
Why do items have no link back to the original post?: Because the feed supplies one that does not resolve. In the window this site was built from, every one of the 1209 items carried the same placeholder URL rather than a permalink to the post it came from. Rather than link every item to a page that is not its source, briefings carry no per-item source link at all. Where an item contains a real URL in its text, that URL is shown and marked nofollow.
Why is there no journalist’s name on the briefings?: Because there is no journalist. The briefings are compiled by the editorial desk from the recorded feed, and the author is given as the desk rather than as a person. The feed’s own author field is not used: in the current window every item carried the same author string regardless of who actually wrote it, so publishing it would attribute many people’s words to one name. Adding a named editor with real credentials would strengthen this site; inventing one would not.
Do you verify what you publish?: No, and the briefings say so on every page. An item is recorded as the feed captured it and is not independently checked against the company’s own announcement. What the site does verify is what it can: entity identities, official channels, and the arithmetic on each page. Claims about products and releases should be confirmed at the official channel linked from the relevant entity page.
How do I get something corrected?: Write to editor@bloger.fm with the briefing date and the item. Corrections that the feed itself got wrong cannot be fixed at the source, so the item is removed from the briefing when a correction is confirmed, and the briefing's modification date changes. Definitions and entity metadata are corrected directly and take effect on the next build.
What does “gaining ground” mean?: It means an entity was mentioned more often in the second half of the window than the feed as a whole grew — so it took a larger share of coverage than it had. It is not a claim that the company is doing well. The comparison is against the feed's own growth because the feed roughly doubled across this window: measured against raw counts, almost everything looks like it is rising, which measures the feed and not the entity. Full method, including the thresholds, is in "What the site works out, and how" above.