Do We Still Need Code Reviews in the Age of Coding Agents?

As coding agents become capable of generating entire features and pull requests, the traditional human-to-human code review process faces a significant shift. Remo Jansen argues that while verification remains essential, the reliance on manual peer reviews for every change is becoming a bottleneck. Instead of treating every pull request as a generic task, the author suggests a more nuanced approach: trusting the engineer who oversees the agent to take ownership, while offloading verification to specialized adversarial agents, automated testing, and robust architectural constraints. By making certain classes of bugs impossible through design, teams can focus human attention on high-level judgment and complex trade-offs. Ultimately, the goal is to increase confidence in software quality by building a layered, composable verification pipeline that balances human expertise with automated efficiency, rather than simply increasing the number of human reviewers in the development cycle.