Google Says AI Writes 75% of Code; Sonar Targets the Verification Gap

LinusEkenstam · x · 2026-08-12

LinusEkenstame points out that while Google claims AI writes over 75% of new code, the measured velocity gain is only about 10%. The gap highlights the massive verification problem created by AI-generated code.

He analyzed Sonar's new tools, Vortex and Hunter Agent. Vortex lives inside coding agents like Claude Code and Cursor, feeding standards before code generation and verifying edits mid-task. Benchmarks show up to 36% lower token consumption and 92% fewer introduced issues. Meanwhile, Hunter Agent acts as an AI security agent that reasons through code like a researcher to catch business logic and access control flaws that traditional automated scanners miss.

Related event: Rise of AI-Generated Code Shifts Bottleneck to Verification(3 posts)→

Original post →

More from coding & agent

coding & agent channel →