AI has dramatically sped up code writing, but this efficiency comes at a cost. The traditional code review process is under strain, not because of large diffs, but because reviewers are overwhelmed by the sheer volume of changes and struggle to determine where to focus their attention. This issue is exacerbated by the fact that AI-generated code often lacks clear intent or scope, leaving reviewers adrift amid a sea of changes. Plans are emerging as a crucial solution. In AI-assisted development, a plan is more than a TODO list; it defines the unit of review, specifying intent, scope, and boundaries. It also serves as the unit of generation, guiding what AI should and should not modify, and as the unit of knowledge, setting the stage for later documentation. By treating plans as first-class artifacts in the code repository, developers can ensure that reviews start from intent rather than diff scanning, shrink PR sizes, and focus on judgment rather than syntax. This approach makes AI reviews more intent-aware, reduces context window usage, and makes future changes more manageable.

The shift towards plans as a first-class artifact in AI-assisted development represents a significant evolution in how teams approach code reviews. It acknowledges that AI-generated code requires a new framework for human oversight. By establishing clear intent and boundaries, plans help developers navigate the complexities of AI-assisted coding. This method not only improves the efficiency of code reviews but also ensures that the human element remains central to the development process, maintaining the quality and coherence of the codebase. The emphasis on plans suggests a future where AI and human developers collaborate more effectively, with AI handling the heavy lifting of code generation while humans provide the critical oversight and decision-making.

As AI continues to transform the development landscape, the role of plans in code reviews is likely to become even more pronounced. Teams that adopt this approach will find themselves better equipped to handle the challenges of AI-assisted coding. The focus on intent and scope will help developers maintain control over their codebases, even as AI tools become more sophisticated. This evolution in code review practices underscores the ongoing need for human judgment and strategic planning in software development. As the industry moves forward, the success of AI-assisted development will increasingly depend on how well teams can integrate these new practices into their workflows.