Skip to content

How the engine works

Receipt

The jobs report (Bureau of Labor Statistics)

The kind of run
Replay of the release of Nov. 20, 2025, run on Nov. 14, 2026
The part of the rule it answers to:
Part 6. Versions
The result
Clean: the story passed the whole gate and matched its answer key.
The release:
2025-11-20

A proving run, never published.

The engine wrote this story from the release of Nov. 20, 2025 as it was first published, to test this kind of story. It is not news and was never on the wire.

Payrolls rose by 119,000 in September 2025; unemployment rate was 4.4%

The Bureau of Labor Statistics also revised its payroll figures for the two months before September 2025, according to its Employment Situation report.


In brief

  • The figure: Total nonfarm payroll employment rose by 119,000 in September 2025, the Bureau of Labor Statistics reported.

  • Unemployment: The unemployment rate was 4.4%.

  • The revisions: The bureau revised the change for July to a gain of 72,000 from a gain of 79,000, and the change for August to a loss of 4,000 from a gain of 22,000.

  • By the numbers: With the revisions, employment in the two months combined was 33,000 lower than first reported.


Total nonfarm payroll employment rose by 119,000 in September 2025, and the unemployment rate was 4.4%, the Bureau of Labor Statistics reported Nov. 20, 2025 (USDL-25-1487).

The bureau revised the change for July to a gain of 72,000 from a gain of 79,000, and the change for August to a loss of 4,000 from a gain of 22,000.

With the revisions, employment in the two months combined was 33,000 lower than first reported.

Written from the Bureau of Labor Statistics’ Employment Situation report for September 2025: the payroll change from its survey of employers and the unemployment rate from its survey of households. Both are estimates from samples; the payroll figures are revised in the next two reports and once a year by the annual benchmark revision. The bureau generally conducts its analyses at the 90-percent level of confidence, at which the confidence interval for a month’s change in payrolls is on the order of plus or minus 136,000: by its own example, a gain of 50,000 could not be said with confidence to be an increase. The gain of 119,000 in September 2025 is within that range. The report also gives the change by industry and the bureau’s own account of it; the story reports the totals and the revisions.


The record it was written from


The facts it used

  • 119,000https://www.bls.gov/news.release/archives/empsit_11202025.htm: “Total nonfarm payroll employment edged up by 119,000 in September but has shown little change since April, the U.S. Bureau of Labor Statistics reported today.”
  • 4.4%https://www.bls.gov/news.release/archives/empsit_11202025.htm: “The unemployment rate, at 4.4 percent, changed little in September.”
  • September 2025https://www.bls.gov/news.release/archives/empsit_11202025.htm, the release’s heading
  • Nov. 20, 2025https://www.bls.gov/news.release/archives/empsit_11202025.htm, the embargo line
  • USDL-25-1487https://www.bls.gov/news.release/archives/empsit_11202025.htm, the release’s number
  • Julyhttps://www.bls.gov/news.release/archives/empsit_11202025.htm: “The change in total nonfarm payroll employment for July was revised down by 7,000, from +79,000 to +72,000, and the change for August was revised down by 26,000, from +22,000 to -4,000.”
  • a gain of 79,000https://www.bls.gov/news.release/archives/empsit_11202025.htm: “The change in total nonfarm payroll employment for July was revised down by 7,000, from +79,000 to +72,000, and the change for August was revised down by 26,000, from +22,000 to -4,000.”
  • a gain of 72,000https://www.bls.gov/news.release/archives/empsit_11202025.htm: “The change in total nonfarm payroll employment for July was revised down by 7,000, from +79,000 to +72,000, and the change for August was revised down by 26,000, from +22,000 to -4,000.”
  • Augusthttps://www.bls.gov/news.release/archives/empsit_11202025.htm: “The change in total nonfarm payroll employment for July was revised down by 7,000, from +79,000 to +72,000, and the change for August was revised down by 26,000, from +22,000 to -4,000.”
  • a gain of 22,000https://www.bls.gov/news.release/archives/empsit_11202025.htm: “The change in total nonfarm payroll employment for July was revised down by 7,000, from +79,000 to +72,000, and the change for August was revised down by 26,000, from +22,000 to -4,000.”
  • a loss of 4,000https://www.bls.gov/news.release/archives/empsit_11202025.htm: “The change in total nonfarm payroll employment for July was revised down by 7,000, from +79,000 to +72,000, and the change for August was revised down by 26,000, from +22,000 to -4,000.”
  • 33,000https://www.bls.gov/news.release/archives/empsit_11202025.htm: “With these revisions, employment in July and August combined is 33,000 lower than previously reported.”

The answer key

The story reports every figure and fact the answer key holds. A program of its own read the key from the release, never the engine’s writer.

  • Found: The monthExpected September 2025. The story says “September 2025”.
  • Found: The change in payrolls, September 2025Expected 119,000 (up). The story says “119,000”, up.
  • Found: The unemployment rate, September 2025Expected 4.4 percent. The story says “4.4%”.
  • Found: The revised change for JulyExpected 72,000 (up). The story says “72,000”, a gain.
  • Found: The revised change for AugustExpected 4,000 (down). The story says “4,000”, a loss.
  • Found: The two revisions togetherExpected 33,000 (down). The story says “33,000”, down.
  • Checked against the bureau’s table of first estimates: September 2025, first estimate 119 thousand (https://www.bls.gov/web/empsit/cesnaicsrev.htm).

The checks

  • Shape: passedHeadline, three to five summary lines, a body, a source and a source note are present.
  • Numbers: passedEvery figure in the text is in the fact sheet.
  • Arithmetic: passedEvery computed figure matches when recomputed, and every verb agrees with its sign.
  • Names: passedEvery name in the story appears in the source or its record.
  • Quotations: passedEvery quotation matches the source exactly.
  • Dates: passedEvery date is one of the fact sheet’s.
  • Source: passedEvery document was read from the issuing body’s own address, stored, and fingerprinted.
  • Freshness: passedThe source is recent enough to be news.
  • Duplicate: passedNo story stands for this event, and the headline is new.
  • Language: passedNo opinion, speculation or first person in the story’s own words.
  • People: passedEvery person named is an official or acts in an office the source names.
  • Rights: passedThe source’s terms allow republication.
  • Answer key: passedEvery one of the 6 figures and facts the release gives is in the story.

The writer and the reviewers

  • Wrote it claude-sonnet-5-5
  • Fact review: passed claude-opus-5-514 claims, each supported by the source.
  • Neutrality review: passed claude-opus-5-5Neutral; the source note says what the source is and what a reader should know about its limits.

The versions

  • Source reader: bls.empsit reader 4.3
  • Writer: claude-sonnet-5-5
  • Reviewers: claude-opus-5-5 (facts); claude-opus-5-5 (neutrality)
  • Instructions: house rules 17.4
  • Checks: twelve checks 17.6, answer key 4.1

The time of each step

  1. Started
  2. Read the release from the agency’s archive
  3. Found the release as new
  4. Kept its copy, with its fingerprint
  5. Built the fact sheet
  6. Written
  7. Put through the twelve code checks
  8. Reviewed
  9. Matched against the answer key
  10. Finished

Run run-20261114-replay-20251120-283926, cycle 6. Every receipt for this kind of story