Senior roles take 71% of the rebound in US software job postings

Put the two Indeed level figures together and the February 2025 starting point was very low. If postings now sit about 27.5% below their pre-pandemic level after climbing nearly 15% since February 2025 ^[1][4], the February 2025 level was roughly 63% of pre-pandemic, or about 37% below it ^[12]. On those numbers the rebound has recovered about a quarter of the lost ground ^[14]. Getting all the way back takes another 38% or so from here ^[13].

The Claude Code launch date also explains less than it appears to. McKinsey Global Institute research says the earlier decline in software hiring was not solely an AI story ^[7]. The rebound comes after hiring fell hard from its 2022 peak, as tech companies retrenched from their pandemic-era expansion ^[7]. An analysis of Bureau of Labor Statistics data by economist Steve Rattner found software developer was among the categories that grew from May 2023 to May 2025 ^[10]. Its growth only matched the average across all occupations, and ran well below blue-collar jobs tied to building AI data centers ^[10]. Eric Kutcher, North America chair at McKinsey, takes the more cheerful line. "We're seeing much more growth in the places we thought we were gonna see the biggest displacement," Kutcher said ^[8].

The mix of openings says more than the total. Between May 2025 and May 2026, senior roles accounted for 71% of the increase in postings and job titles mentioning AI for 37% (the two groups may overlap, so they cannot be added) ^[5][6]. Of every 100 postings added, 29 went to roles that were not senior ^[15]. CNBC's report says faster AI output could concentrate demand on workers who can evaluate AI-generated code and connect new tools to a company's wider systems ^[11]. Sneha Puri, an economist at Indeed Hiring Lab, said companies still need developers to oversee AI's coding work, refine products and help clients fit the technology into daily operations ^[9].

The code that needs overseeing is growing much faster than the postings. GitHub said in October that one in three pull requests now involve an AI agent, against fewer than one in 10 a year earlier ^[2]. That is more than a threefold rise in twelve months, set against a postings gain of nearly 15% over a longer stretch ^[16][1]. "If that pace holds, within the next two years, most of the code pushed to GitHub could be written by an agent. Much of it may never be fully read by a human," GitHub stated ^[3]. GitHub has not confirmed claims from some AI experts that AI-written code across the platform has already passed half ^[17].

The evidence fits more than one outcome. In one, the rebound is mostly the hiring cycle and stalls well short of pre-pandemic levels whatever agents do ^[7]. In another, agent share keeps rising on GitHub's curve, review becomes the scarce input and senior postings climb further ^[3]. The third has employers buying seniors while keeping junior intake thin, drawing down the pool they hire from ^[15].

In our view the second is the plan the data supports for the next year or two, and employers are already acting on it: senior roles took most of the increase ^[5]. What they are not doing is opening junior seats at the same pace ^[15]. The counter-thesis is the first outcome, in which the Claude Code date is a coincidence and postings would have risen with the cycle anyway ^[7]. We would be wrong if senior postings flatten while the agent share of pull requests keeps climbing. That combination would mean companies are accepting code nobody reads instead of paying someone to read it.