Summary: Since Anthropic launched Claude Code in February 2025, US software developer job postings have climbed nearly 15% while overall postings fell 7%, per Indeed Hiring Lab data reported by TechOrange on October 12, 2026. AI is writing more code than ever, yet demand for engineers who can direct and review it is growing. The catch: that demand skews heavily toward senior and AI-focused roles.
The numbers seem contradictory
GitHub reported in October that one in three pull requests now involves an AI agent, up from under one in ten a year earlier. GitHub itself said that if current rates hold, most code submitted to the platform could be agent-written within two years, and much of it may never be fully read by a human. Some AI experts claim AI already writes over half the code on GitHub; GitHub hasn’t confirmed that.
Meanwhile, the job market moved the other way from predictions. Indeed Hiring Lab’s data shows developer postings up nearly 15% since February 2025, against a 7% decline in overall postings. As McKinsey’s North America chairman Eric Kutcher put it on CNBC’s Squawk on the Street, companies are producing more code than ever, and developers who can work with AI tools remain in demand — growth appeared exactly where people expected the most displacement.
What kind of demand is actually growing
The recovery is uneven, and the composition matters more than the headline. According to figures cited from Indeed Hiring Lab, postings are still about 27.5% below pre-pandemic levels. Between May 2025 and May 2026, senior roles accounted for 71% of the increase in developer postings, and roles with AI in the title made up 37% of it.
Indeed Hiring Lab economist Sneha Puri’s explanation is practical: even when AI handles technical coding tasks, companies need developers to supervise the work, keep improving products, and help customers integrate these tools into daily operations. Demand is rising for people who can put AI to work inside an organization, not just build the tools. That maps to what I flagged earlier when Anthropic’s frontier coding models started shipping at mid-tier prices — the bottleneck shifts from writing code to routing, reviewing, and owning it (see frontier coding output priced like a mid-tier model).
Why this might not last
Both Anthropic researchers and GitHub flag the fragility here. Anthropic’s team noted that human oversight of automated work may not persist as the pattern. GitHub’s own report is blunter: as AI brings more code into the pipeline, potential vulnerabilities scale with it, and without mechanisms to review AI output efficiently, the volume of newly introduced security issues becomes unmanageable.
There’s also a generational split. Stanford’s 2026 AI Index Report found that since 2024, employment for 22-to-25-year-olds in the most AI-exposed occupations ran about 16% lower than in the least exposed — and early-career software developers specifically declined nearly 20%, even as senior roles expanded. Economists also caution that the earlier downturn wasn’t purely AI-driven; it followed the 2022 post-pandemic hiring peak and correction.
What builders should take from this
For anyone hiring or planning a career: the market is pricing judgment, not typing speed. Puri’s advice to students was to get fluent with AI tools but keep investing in communication, interpersonal, and leadership skills — the parts of the job that don’t compress. She also put an obligation on employers: today’s junior talent is tomorrow’s senior expert, and companies that stop building entry-level pipelines won’t have the senior people they need in a decade.
The US Bureau of Labor Statistics’ August report still projects 10% employment growth for software developers from 2025 to 2035, from roughly 1.72 million to 1.89 million positions. That’s a reasonable baseline — but treat it as a forecast about a transforming role, not a guarantee that today’s entry path stays open.