AI now handles code review's tedious work

What happened

A technical opinion piece argues that human reviewers should use AI to scan large code changes (10,000+ lines) rather than reviewing them manually line-by-line, because LLMs have become skilled at catching high-severity vulnerabilities and are better suited to nitpicking than human reviewers.

Why it matters

Code review is a bottleneck in many engineering teams, but human reviewers' real value lies in catching design problems and architectural inconsistencies that only someone with deep codebase knowledge can spot — knowledge that neither the code author nor the LLM possesses, such as deprecation plans discussed in recent meetings or team-specific coding conventions.

What to watch

The advice comes with a caveat: this workflow applies to codebases where individual lines are not critical. In domains like embedded systems, where each line may be sacred, traditional line-by-line human review remains necessary.