Google now has AI write three-quarters of its new code, with humans reviewing afterward

Google announced that 75 percent of new code written at the company is now generated by AI (likely using its Gemini or internal code-generation tools), then checked and approved by human engineers before being used.

Unlike older code-suggestion tools that offer one-line completions, Google's AI writes larger blocks of code that humans must validate for correctness and security — shifting the developer's role from writing to reviewing and editing.

For software engineers at Google and elsewhere, this means the bottleneck moves from 'writing code from scratch' to 'making sure AI-generated code is safe and does what it should' — a skill shift that could reshape hiring and training across the tech industry.