tokens agents

State of the market · Swept Aug 31, 2026, 07:30 UTC

The first sweep, in full.

Every figure here is a reading. It came either from one night of reading company job feeds direct, or from research we name in full underneath it. Nothing is modelled and nothing is projected. Where we hold one observation, the chart shows one observation and says so on its face.

The headline number

0

postings caught by a named rule this sweep

Five rules ran. Every one of them reports what it found, including the ones that found nothing, because a rule missing from a summary reads as a rule nobody ran.

This is the only number on this page we produced ourselves, and it is a reading rather than a fact: it is what the machine measured last night and it will be a different number tomorrow. Everything else here is somebody else's research, cited where it sits.

The figures

Figure 1 · Ranked bar

What each rule found

  • repost churn

    0

    A req that left the boards and came back under a new identifier, freshly dated, with its title, location, remote flag, employment type, posted pay and description byte for byte the ones it had before.

    It compares tonight against closed records carrying a content fingerprint, and the machine only started writing that field on August 19. It reports nothing until the archive has the records to compare, and that zero is a result rather than a rule missing from the page.

  • misrepresented

    0

    The location the employer printed offers remote and nothing else, and the posting's own description states an in-office requirement, with no remote option anywhere in its text.

  • zombie

    0

    The posting is still listed past a closing date the employer printed themselves, with an explicit four digit year beside it.

  • phantom

    0

    The posting's own link is gone at source: a 404 or 410, or a redirect away from the posting's own identifier, confirmed on a second request.

    Ashby answers 200 for postings that do not exist, and Ashby is about four fifths of the boards we read, so phantom cannot be established there at all. A rule that cannot see is told to say nothing rather than to guess.

  • touched, not refreshed

    0

    The published date moved while the title, location, remote flag, employment type, posted pay and description all stayed byte for byte identical.

Every rule ran and every rule found nothing. That is the whole result, printed rather than summarised away. No bars are drawn beside these counts, because a track of length zero reads as a measurement of zero and this is an absence of findings.

Each rule states what it observed rather than what it concluded. The counts are readings from the sweep, so they are as current as the night they were taken and no more.

Measured by The Index, swept Aug 31, 2026, 07:30 UTC

40% of hiring managers admit posting a job they never intended to fill

Context: ResumeBuilder survey

That survey asks hiring managers what they did. It is the only honest source for intent, because a machine reading job feeds cannot observe one. Nothing above it is a measurement of intent, and the two are not two views of one number.

Figure 2 · Distribution

When applications arrive

  • 45% In the first 48 hours
  • 60% By 96 hours, running total
  • Not published After 96 hours

45% of applications arrive within 48 hours of a posting going up, and 60% within 96 hours

Source: Davis and Samaniego de la Parra, NBER Working Paper 32320, 125 million applications

Median posting life is 7 days; only a quarter stay up past two weeks

Also: Davis and Samaniego de la Parra, NBER Working Paper 32320, 125 million applications

Figure 3 · Ranked bar

The sweep, stage by stage

  • Company feeds read 45
  • Postings pulled 66
  • Verified live at source 66
  • Killed by rule 0

Four stages, one night. Each bar is a count the machine recorded as it went, not a total anybody reconciled afterwards.

Measured by The Index, swept Aug 31, 2026, 07:30 UTC

The first bar counts company feeds and the other three count postings. They share an axis of counts, not of like things. The feed count stays because it is the one stage a reader cannot work out from the others.

Figure 4 · Hero number

What one posting attracts

254

applications on the average job

The average job now draws about 254 applications

Source: Greenhouse, via Fortune, July 2026

49% of hiring managers name AI-generated sameness a top red flag

And: Resume Genius, 1,500 hiring managers, August 2026

How these figures were made

Facts and readings
Two kinds of number, kept apart. A citation is somebody else's research: static, quoted word for word, with its source under it. A reading is ours: what the machine measured last night, computed when this page was built and true only of that night. Nothing on this site states a reading as a fact, which is what our own kill count used to be doing.
Read direct
Every count on this page came from a company's own feed, read at source on the night of the sweep. No aggregator sits between a posting and this page.
One clock
Every time on this page is the sweep instant, Aug 31, 2026, 07:30 UTC. There is no second clock and no time of rendering anywhere on this site.
Nothing modelled
No projection, no smoothing, no line drawn between two published points. Where a number was not published, the chart shows the gap rather than filling it.
Every figure sourced
Each figure names who measured it, in the machine voice, directly under the chart. A figure with no source line fails the build before it can reach a reader.
A closed fact list
The statistics this site may state live in one file with their sources, and this page renders six of the seven. The one it leaves out is Greenhouse's activity figure, which is licensed for this page and stays off it: the only place it would sit is beside our own counts, and the two do not measure the same thing.