Code review is dead. Long live intent verification and the AI slop

Ankit Jain argues that traditional code review is breaking under AI-generated code volume. With code churn up 861% and 30% of changes merged without review, the old model of line-by-line human review is already dead in practice. The real crisis is that AI writes code and AI reviews it, leaving humans as a

superficial validation layer. The blind spot is that code reviews serve a deeper purpose: alignment, knowledge sharing, and architectural decision-making. Jain introduces a two-part framework. Semantic accuracy (catching bugs, security issues) can be automated via an "AI slop registry" that codifies recurring review comments into guardrails. But alignment—the collaborative understanding of intent, design

decisions, and tradeoffs—must survive. This is the part most teams neglect when rushing to automate reviews. The proposed solution shifts the review surface from diffs to intent verification, capturing prompts and session decisions as acceptance criteria. A verification system then tests behavior against those criteria, producing evidence like screenshots and database snapshots. Reviewers

examine intent and evidence, not code. This moves discussions from syntax to architecture, which is where senior engineers add value. The approach mirrors behavior-driven development but with AI-generated test plans from session data. The key tradeoff is upfront investment in building the registry and capturing intent, which follows a J-curve before paying off.