The Bureau of Labor Statistics dropped the April 2026 JOLTS release at 10:00 ET on Tuesday, and the headline number is the kind that gets cable-news chyrons excited. Job openings climbed to 7.62 million, the highest level since July 2024, a 731,000 jump from March — the biggest single-month rise since 2021. A casual reader sees that line and assumes the labour market just turned a corner.
Three rows down, hires fell by 419,000 to 5.12 million. The hires rate dropped to 3.2%. Total separations were little changed at 5.0 million. Quits sat at 3.0 million for the eighth straight month. Layoffs and discharges came in at 1.7 million, with the rate ticking down from 1.2% to 1.1%. The four numbers, read together, describe a labour market in which postings ballooned, hiring collapsed, firing flatlined, and quitting refused to budge.
Indeed Hiring Lab’s morning note captured the disconnect with the line “the bigger they are, the harder they hire.” The phrase economists have been using for fifteen months — low-hire, low-fire — is no longer a transitional description of a recovering market. It is the steady state.
What the four numbers actually say
Postings up. Hiring down. Firing flat. Quitting flat. This is not a recovering economy. It is an economy in which the funnel between a posted job and a filled job has narrowed to the width of a straw.
The 2022 reference month is useful here. At the May 2022 peak, the US had 12.0 million openings, 6.7 million monthly hires, and 4.5 million monthly quits. In April 2026, the country has 7.62 million openings, 5.12 million hires, and 3.0 million quits. Openings are 36% off their peak. Hires are 24% off. Quits are 33% off. The economy is posting two-thirds of its 2022 demand, hiring three-quarters as many people for those openings, and watching workers move between jobs at the slowest non-recession pace since 2014.
The standard interpretation — businesses cautious, workers cautious, both sides waiting out a rate cycle — does not fully cover this. Rates have been falling. The Fed cut in March and is signalling another cut by September. The waiting-out-the-Fed story would predict openings and hires moving together. They are not. Postings expanded; hiring shrank. The two sides of the funnel are decoupling.
Why a posted job is no longer an offered job
A posted job in 2022 was a budget commitment and a recruiter’s KPI. A posted job in 2026 is increasingly a job description published to satisfy procurement or compliance and then quietly frozen. The recent run of “ghost job” reporting — postings left up to harvest CVs, build talent pipelines, signal growth, or comply with policies that require external posting before internal hire — explains some of the gap. AI tools that auto-generate JDs make posting nearly free; staffing budgets that have to clear a CFO desk make filling expensive.
The second mechanism is the one Stanford’s AI Index economy chapter put a number on in April: companies are posting roles they used to hire for and then either freezing them or quietly absorbing the work with AI tooling. The April Indeed Hiring Lab note flags large-firm hiring as the slowest tier of the entire labour market — exactly the cohort most likely to have an internal AI productivity program with explicit headcount-avoidance language. The 22-to-25 software developer cohort that Stanford clocked at minus 20% since 2024 is the canonical example. The reqs are open. The offers are not extended.
The third mechanism is the AI-rebalanced posting. A company cuts 4,000 engineers in May, opens 4,000 reqs in June, and waits to fill them with senior AI-platform hires it cannot actually find. The posting goes into JOLTS. The hire never materialises. The req sits open at month-end, which counts as another opening, which keeps the headline number elevated.
The quits floor is the worker half of the same story
The quits rate of 1.9% is the lowest non-pandemic reading since 2014. Workers are not quitting because they cannot tell which side of the next layoff list they are on, and because the job-finding rate for those who do quit is the weakest it has been outside a recession in a decade. The combination — bosses afraid to fire experienced workers and workers afraid to leave the seat they have — is the “low-hire, low-fire” floor in human-emotion terms.
What is new in April 2026 is that the openings number jumped while the quits floor held. That is not a sign of confidence returning. It is a sign of postings inflating without changing the underlying hiring or movement dynamics. The same labour market is now harder to enter, harder to leave, and visibly easier to appear to be hiring into.
The two numbers that should be read together
JOLTS came in the same week the TrueUp tracker logged 148,173 cumulative tech displacements for 2026 — 355 events at 975 a day, 46% above 2025’s pace. The same month the BLS counts 731,000 more openings than March, the tech-cuts ticker keeps running. Both can be true because they are different funnels. Openings are gross postings across the whole economy. Tech cuts are realised layoffs in one sector. The two are not contradictions — they are the two halves of the new normal: lots of postings, lots of cuts, very little net movement in between.
The closer comparison is openings against actual hires. April posted a 7.62M-to-5.12M ratio: one hire for every 1.49 openings. In May 2022, the ratio was 12M-to-6.7M: one hire for every 1.79 openings. So the funnel was actually tighter at peak demand. The April number is concerning not because the ratio shifted dramatically, but because openings can grow 11% in a single month while hiring drops 8% — and nobody at the BLS press call could explain why.
What the May print will tell us
May JOLTS is due July 1. Three things to watch. First, whether the April openings spike sustains or reverses — a one-month bounce attributed to seasonal adjustment glitches is a different story than a new floor. Second, whether the hires rate stays at 3.2% or finds a way back to the 3.4% it held through Q1. If hires keep falling while openings stay elevated, the decoupling becomes the headline rather than the footnote. Third, whether quits move at all — a quits rate ticking up would be the first concrete sign that workers think they can move; a quits rate stuck at 1.9% confirms that they don’t.
The Atlanta Fed’s GDPNow is currently tracking Q2 growth at 2.4%. The labour market is supposed to look like an economy growing at 2.4%. It looks instead like an economy posting jobs at 2026 levels while hiring at 2014 levels and watching workers move at a recession pace. That is the gap the JOLTS release just made unignorable.
The structural reading
The simplest model that fits the data is also the most uncomfortable. A meaningful share of US labour demand has shifted from “we need a human in this seat” to “we have budget for the seat and an AI tool that might fill it before we have to.” Postings reflect the budget. Hires reflect the resolution. The gap between the two is the share of demand that is being absorbed by software rather than people.
The April print does not prove that thesis on its own. It is consistent with it. So is Stanford’s 20% junior-developer collapse, so is Cisco’s 4,000 cuts on the same day as a record $15.8B quarter, so is the TrueUp 975-a-day tech-cuts pace, so is Paul Osterman’s May 31 reminder that “AI” has been the cover story for two decades. The numbers don’t all point in exactly the same direction. They point in directions that are no longer in tension.
The thing JOLTS made plain on Tuesday is that the openings-up-hires-down story is not a quirk of one sector or one quarter. It is now the shape of the whole US labour market. The seat is open. The offer is not coming. The worker who would have filled it is on a tracker, and the worker who would have quit it is sitting still.
