Verification — not code generation — is now the bottleneck

The theme converging across Hacker News threads and June writeups: code volume is no longer the constraint; review capacity is. Sonar's 2026 State of Code survey found 96% of developers don't fully trust AI-generated code to be functionally correct, yet only ~48% always review it before committing — and PR review time rises 91% when teams adopt AI assistants even as individual task completion climbs.

The practical response is to treat verification as a first-class discipline and invest in review throughput the way you'd invest in build speed: smaller PRs, automated quality gates (static analysis, security scanning) on AI-generated changes, and human review reserved for architecture and business logic. "Vibe, then verify" — accept fast, validate systematically.

Sources: DEV: The Verification Gap, byteiota: 96% don't trust AI code, Hitechies: the best engineers aren't the best coders