Researchers propose Adversarial Review (AR), a protocol for cooperative code review where three agents work together: a main coding agent, a reviewer, and a critic. The critic audits the review through structured disagreement before the main agent edits. Experiments on various benchmarks show that AR outperforms other methods in terms of pass rates and F1 scores. --- Why it matters: This matters to engineers because it shows that cooperative code review can be effective even with minimal agentic cooperation, which could lead to more efficient and accurate coding tasks. Source: https://arxiv.org/abs/2608.18167