The NYC AI Hiring Index

What the index measures

The cuts below are all derived from the same daily reading. Each one is computed at collection time, because a posting that has been filled cannot be re-read afterwards — there is no archive of job boards to go back to.

How we count, and what we refuse to do

The short version is below. The full methodology covers every series, its known limits, and a dated log of each change we have made, and errors are recorded in the public corrections log.

Changes are like-for-like, always

Adding a company to coverage adds its entire board to the total. Differencing the headline number would therefore report our own bookkeeping as a hiring surge. Every change figure we publish is computed across the companies present in both periods, and we print the headline difference next to it so you can see the gap. In the first issue that was the difference between an exciting +2.9% and an honest +0.4%.

A partial outage is not a hiring cliff

If fewer than 80% of the boards respond on a given morning, the run records nothing at all. An outage and a collapse in demand look identical in the data and only one of them is real. Closures are only evaluated for companies whose board actually answered that day.

Small cells stay empty

Any salary cut with fewer than eight observations is published as blank rather than as a number with a wide invisible error bar. The same applies to every other series: two closures is an anecdote, and we will say so rather than calculate a median from it.

Traffic goes to the employer

Every listing links to the company’s own application page, never to a form of ours. We are not a recruiter and we are not standing between you and the employer.

A series waits until it supports a number

Per-company hiring velocity needs roughly eight weekly observations before a move can be called unusual rather than normal variation — in the first week, only 12 of 46 companies moved at all and the median move was a single role. Time-to-fill waited the same way for closures to accumulate, and is reported as a cohort in the weekly issues. A new series does not appear here until the data supports it.

Who is in the index

Coverage is deliberately narrow and named. We would rather read every board of 47 companies each morning than sample a thousand and tell you it is the market. A company enters coverage when it is headquartered in New York or runs a substantial New York office, builds AI as the product rather than buying it, and publishes a machine-readable job board.

When a company joins, its whole board joins with it — which is exactly why we never difference the headline total. The two columns below are the distinction that most hiring coverage misses: a large board is not the same as a large New York presence.

Reading of November 14, 2026. A company missing from this list is missing because it has no public board we can read, not because we judged it unimportant — tell us who to add.

Who reads this, and what they use it for

Using and citing the data

Every figure published here may be quoted, charted, or built on, in commercial work as well as editorial, with attribution to the NYC AI Hiring Index and a link to this page. We would rather the number travel with its method attached than travel alone.

When you cite a figure, cite the reading date with it. This is a live series, not a static report: the total changes every morning, and a number quoted without its date will be wrong within a week through no fault of yours.

Need a cut we do not publish — a single company’s history, a function we do not break out, or the underlying series for a period? Ask. Requests from journalists and researchers are answered free.