Summary: AI now writes a large share of the code flowing through GitHub, yet US software developer job postings have grown nearly 15% since February 2025 while overall postings fell 7%, per TechOrange’s October 12, 2026 report citing Indeed Hiring Lab and CNBC. The catch: almost all the growth is in senior and AI-focused roles. For builders, the takeaway is that reviewing, integrating, and directing AI output — not typing every line — is where teams are spending headcount.
The numbers, side by side
Two trends are moving up at the same time. GitHub’s October data shows one in three pull requests now involves an AI agent, up from under one in ten a year earlier, and GitHub says most code committed to the platform could be agent-written within two years if the pace holds — possibly never fully read by a human. Meanwhile, TechOrange reports, citing Indeed Hiring Lab, that US software developer postings are up nearly 15% since Claude Code launched in February 2025, against a 7% drop in overall postings.
Those facts look contradictory only if you assume AI-generated code replaces engineers. The market is pricing something else: oversight.
Why demand shifted instead of shrinking
McKinsey’s North America chair Eric Kutcher told CNBC’s Squawk on the Street that companies are producing more code than ever and developers who can use AI tools remain in demand — growth, he noted, is appearing exactly where replacement was expected to hit hardest. McKinsey Global Institute also points out the earlier downturn wasn’t purely AI-driven; it followed the 2022 hiring peak and post-pandemic pullback.
The composition of new postings matters more than the total. Per Indeed Hiring Lab, senior roles accounted for 71% of the increase between May 2025 and May 2026, and roles with AI in the title made up 37%. Economist Sneha Puri’s framing, as reported by TechOrange: companies need people to supervise AI work, keep improving products, and help customers integrate the tools into operations — applying AI, not just building it.
What this means for how you build and hire
If most of your codebase is heading toward agent-written territory, your leverage moves to three places: evaluating generated code before it merges, wiring new tools into your existing systems, and specifying problems well enough that agents produce usable output. That’s a different job description from two years ago, and the posting data suggests employers already know it. If you’re weighing coding-agent costs against throughput, the economics of routing work across model tiers is worth a look — I covered how frontier coding output got repriced like a mid-tier model in an earlier post on cost per task.
The caveats worth keeping in view
This growth may not hold. Even after the rebound, developer postings sit about 27.5% below pre-pandemic levels, per Indeed Hiring Lab. GitHub itself warns that rising code volume raises security exposure and that human oversight mechanisms must scale accordingly — otherwise the count of newly introduced vulnerabilities becomes unmanageable. Anthropic researchers, per TechOrange, note that whether the current human-supervision pattern persists is “far from certain.”
The people most at risk are the youngest. Stanford’s 2026 AI Index found that since 2024, employment for 22- to 25-year-olds in the most AI-exposed occupations sits roughly 16% below the least exposed — and early-career software developers dropped nearly 20% while senior roles kept expanding. Puri’s advice to students: get fluent with AI tools, but keep the communication and leadership skills that still matter, and employers need to keep entry pipelines open, because today’s juniors are tomorrow’s seniors.
The US Bureau of Labor Statistics still projects 10% developer employment growth from 2025 to 2035, adding about 174,700 jobs. That’s a reasonable baseline — but plan around the shape of the demand, not just its size.