Artificial-intelligence coding assistants and a fresh wave of layoffs at companies such as Amazon and Microsoft are eroding the entry-level software jobs that once drew new graduates into the tech industry. Around 170,000 students were enrolled in U.S. computer-science programs in 2024—double the level a decade earlier—but the unemployment rate for recent graduates has climbed to roughly 6%, compared with about 3% in fields such as biology or art history. Many newcomers, including California-trained coder Manasi Mishra, say the only interviews they receive now are from fast-food chains, a stark shift from the $165,000 starting salaries that were common just a few years ago. Recruiters and educators warn that the tasks most easily automated—documentation, bug fixes and basic feature work—are precisely those that serve as on-ramps for junior engineers. “Those positions that are most likely to be automated are the entry-level positions,” one careers advisor said, adding that students who fail to master AI tools risk being screened out before a human ever reviews their résumés. At the opposite end of the market, demand for elite AI researchers is intensifying. Anthropic, the San Francisco-based large-language-model developer, has hired the three co-founders of London-born startup Humanloop—chief executive Raza Habib, chief technology officer Peter Hayes and chief product officer Jordan Burgess—along with about a dozen engineers and researchers. Although the terms were not disclosed, the move follows similar acqui-hires across the sector and underscores how intellectual property increasingly “lives in the brain,” as one Anthropic executive put it. The twin trends—vanishing entry-level roles and bidding wars for senior specialists—highlight a widening split in the tech labour market. Analysts say the shake-out is likely to continue as generative-AI systems spread deeper into corporate workflows, forcing universities to update curricula and graduates to acquire skills that machines cannot yet replicate.