Methodology
PlainRecord publishes short accounts of documented events. Every brief is assembled by software from a structured public record and worded using a fixed template written and reviewed by a person in advance.
Clear accounts of documented events, with visible sources, explicit limitations, and transparent corrections.
What this is not
This is not a claim to truth, and not a claim to be free of bias. Choosing which sources to read, which records within them are eligible, and which words describe them are all editorial choices, and they shape what you see here. What is claimed is narrower and checkable: every figure can be traced to a named record, and every change is shown.
How a brief is made
- 01 Fetch. A worker polls a small number of approved endpoints.
- 02 Preserve. The record is stored exactly as received, with the URL it came from, the time it was retrieved and a checksum.
- 03 Normalise. The fields this service uses are extracted. Each one is recorded together with the field it came from or the calculation that produced it.
- 04 Validate. A record that is malformed, incomplete, internally contradictory, or of a kind this service does not report is held and not published. Held records are reviewed by a person.
- 05 Render. Validated values are placed into a fixed template.
- 06 Publish or hold. Only records that pass every check are published.
No language model writes anything you read
No generative model runs when a brief is produced, checked or published. The wording comes from templates that are fixed in the source code and versioned; each page states which template version produced it. Software that generates text was used while building this service, and any template it proposed was reviewed and approved by a person before use.
Selection rules
European Court of Human Rights judgments
Chamber and Grand Chamber judgments of the European Court of Human Rights, in English. Committee judgments are not reported: they apply well-established case-law to repetitive applications and are roughly two thirds of the Court's output by volume, so excluding them is a large omission and is stated here rather than left to be discovered. Decisions on admissibility, communicated cases, advisory opinions and legal summaries are not reported either. Where the Court found a violation on one limb of an article and none on another, that is reported as appearing in both findings; neither half is ever stated without the other. Case names are reproduced as the Court publishes them, including where they name an applicant and where the Court has anonymised one. This service makes no separate anonymisation decision. A judgment establishes what the Court decided on a date. It does not establish anything about the underlying events beyond the Court's own findings, it does not report what either party argued, and a Chamber judgment may still be referred to the Grand Chamber.
European Parliament plenary votes
Final plenary votes on a text as a whole, where Parliament's own record identifies the document explicitly and reports both an outcome and aggregate counts. Amendment votes, split and separate votes, procedural motions, votes taken by show of hands without recorded counts, and sittings where no single whole-text vote can be identified with certainty are all excluded, and the excluded cases are the majority. How individual members voted is discarded at ingestion and never stored, so this service cannot answer questions about it. A document's title is reproduced as Parliament's title, to identify the record; it is not this service's description of the subject, and a vote does not establish that anything the document says is correct.
Eurostat HICP annual inflation
All-items HICP annual inflation on the ECOICOP ver.2 classification, for nine EU reporting areas. Every figure is the one Eurostat published. Two geography labels are shortened so that a headline is readable — Eurostat's own label is shown on the brief itself in every case, and no code or figure is altered. Month-on-month differences are given in percentage points, never in percent, because the figures are already rates. A comparison is omitted entirely whenever Eurostat flags either month as a break in series or a change of definition. Provisional and estimated flags are carried into the wording. No explanation of why a figure moved is offered, and none should be inferred from the ordering of these pages.
Eurostat police-recorded offences
Offences recorded by the police in the twenty-seven EU member states, for eight offence categories: intentional homicide, serious assault, sexual violence, robbery, burglary, unlawful acts involving controlled drugs, fraud, and acts against computer systems. Eurostat publishes twenty-five categories; the other seventeen are not reported. The figures are annual and arrive about a year and a half in arrears, so this section is quiet for most of the year and then publishes a year's worth at once. Recorded offences are not offences committed. The count moves with how willing people are to report an offence, with how the police record what is reported, and with changes in law and counting rules, as well as with offending itself. A rise or a fall does not by itself establish that any of those has changed, and no brief here says it does. Eurostat publishes what each country supplies and does not harmonise recording practice between them, so figures for different countries are not comparable with one another. No brief here places one country beside another: a comparison is only ever a country with its own preceding year, from the same series and the same category, and it is withheld entirely where Eurostat flags either year as a break in series or a change of definition. Differences between years are given as a number of offences. No percentage change is published: on a small base a proportional figure reads as a large movement when the underlying difference is a handful of records.
Eurostat unemployment rate
Harmonised, seasonally adjusted unemployment for the total population aged 15 to 74, as a percentage of the labour force. A filter that does not match the approved definition causes the record to be held rather than quietly substituted. Two geography labels are shortened so that a headline is readable — Eurostat's own label is shown on the brief itself in every case. Month-on-month differences are given in percentage points. This measures unemployment as the Labour Force Survey defines it, which is not the same as the number of people registered as unemployed.
GDACS disaster alerts
Orange and red alerts from the Global Disaster Alert and Coordination System: floods, tropical cyclones, droughts, wildfires and volcanic activity. Green alerts are not reported — they are the great majority of what GDACS emits, eighty-five wildfire alerts in a single week in the week this source was built, and green means little or no humanitarian impact is expected. Earthquakes are not reported from GDACS at all: GDACS takes them from USGS, which this service already reads directly under its own magnitude rule, and publishing both would put one earthquake on the site twice from two records that can diverge after revision. An alert level is not a measurement of the hazard and not a count of anyone harmed. It is GDACS's automatic estimate of the humanitarian impact a hazard of that size is likely to have, calculated from the hazard's extent and the population and vulnerability of the area, and published within minutes of a detection. No brief here reports deaths, injuries, displacement or damage, because the record used does not establish them. A severity figure is published only where GDACS supplies a unit whose meaning has been checked. Floods and volcanic alerts arrive with a severity of zero and no unit at all, so no severity is shown for them and the brief says why. GDACS's own sentence describing a severity is never reproduced: for tropical cyclones it pairs a storm classification with a wind speed that contradicts it. GDACS states that its information is indicative, that it should not be used for decision-making without other sources, and that it does not replace any warning issued by a national or local authority. That statement appears on every brief.
USGS earthquakes
Reviewed earthquakes of magnitude 6.0 and above, worldwide. Only magnitude, location, depth and time are reported, together with any later revision USGS makes. Automatic solutions USGS has not yet reviewed, events it classifies as something other than an earthquake, and anything below magnitude 6.0 are not reported. Effects, damage and casualties are never reported, because the USGS record used here does not establish them.
Times
Four different times are kept apart, and briefs say which is which: when the event happened, which period a figure covers, when the source last revised its record, and when this site published or changed the page. All times are shown in UTC and labelled.
What is never inferred
- Casualties, damage, or the effects of an event.
- Causes, motives, or intentions.
- The consequences of a policy or a vote.
- That a document's contents are true because an institution voted on it.
- That a comparison is valid when the definitions, geography, units or periods behind two figures differ.
Percentages and percentage points
Inflation and unemployment figures are already rates expressed in percent. The difference between two months of such a rate is measured in percentage points, never in percent, and briefs say so explicitly.
Quiet days
When no record meets the rules, the relevant section stays empty. Nothing is substituted to fill space.
Corrections and revisions
Two different things are recorded separately. When a source revises its own record, the brief is updated and the change is labelled as an upstream revision. When this service makes a mistake, that is labelled as our error. Both appear on the brief and on the corrections page.
Coverage gaps
Coverage is narrow by design and the gaps are substantial. They are listed on the source coverage page.