Google CEO Sundar Pichai disclosed in April 2026 that 75% of all new code at Google is now generated by AI and reviewed by human engineers. That figure stood at 25% in late 2024 and 50% by the end of 2025. In under 18 months, AI went from a writing assistant to the primary producer of code at one of the world's largest technology companies.
Microsoft, Meta, and Snap are all moving in the same direction. Microsoft's CTO has stated he expects 95% of all code to be AI-generated within five years. Meta has set internal targets for the first half of 2026 where 65% of its engineers are expected to use AI to write more than 75% of their committed code. The direction is consistent across the industry.
The jobs data tells a blunt story. Stanford researchers found that employment among software developers aged 22 to 25 fell nearly 20% from its late 2022 peak by July 2025. Big tech companies cut entry-level hiring by more than 50% over three years. Job postings for software development roles were still 36% below February 2020 levels as of late 2025. Junior and mid-level roles took the biggest hit: junior tech postings dropped 34% compared to five years ago, and senior postings dropped 19%.
Here is what is actually happening. AI tools now handle the routine work that companies used to hire young engineers to do: writing standard code, fixing bugs, building documentation, running tests. That entry-level work was also how engineers built experience and moved up. That pipeline is under serious strain. The concern, as one developer framed it plainly, is that if nobody gets hired as a junior in 2025, there may not be enough experienced engineers to review AI-generated code a decade from now.
But not everyone is losing. Research published in Science, examining over 30 million code contributions across six countries, found that AI tools boosted productivity and accelerated career growth, but only for senior-level developers. Entry-level developers used the tools the most, yet did not appear to benefit. Engineers who can design systems, catch AI errors, evaluate security risks, and make judgment calls about whether an AI-written solution is actually correct are in a stronger position than before. Workers with multiple AI skills earn a salary premium estimated near 43% above non-AI counterparts.
Also worth noting: AI-generated code is not clean. Research cited by security analysts found that AI-generated code contains nearly three times more security vulnerabilities than human-written code. Gartner forecasts that by 2028, a quarter of enterprise data breaches will trace back to AI agents. The demand for people who can catch those problems is real.
Tech workers are responding with collective action at a pace the industry has not seen before. Over 1,600 Meta employees signed a petition against a company program to collect their computer activity for AI training. More than 600 laid-off Oracle workers signed a joint letter requesting higher severance and longer healthcare. Google DeepMind workers in the UK voted to unionize. In a 2024 survey of about 1,900 tech professionals, 67% said they would be likely to join a union if their company had one.
For anyone running a business that depends on software, the key implication is practical: AI-generated code is already inside your systems. The question is who is reviewing it and whether they are skilled enough to catch what the AI got wrong. The profession has not disappeared. It has split into two tiers, those who can govern AI-generated work and those who cannot, and the gap between them is growing fast.
This pattern is not limited to engineers. The same split is already visible in finance, legal support, HR, and marketing: AI absorbs the routine bottom layer, the oversight and judgment layer holds or grows, and the people in the middle face real pressure. Software engineers arrived there first. Other professions are following.