AI broke code review. What about knowledge sharing?

The volume of AI-generated code has overwhelmed code review, with engineers now reviewing lengthy diffs they didn't write, at a pace that makes careful reading impossible. Vanitha Kumar, Market Technology Director at ThoughtWorks, accidentally discovered a solution: shifting code review left, which means knowledge sharing should also shift left.

Code review has always been a platform for knowledge exchange, allowing junior engineers to learn from senior engineers and fostering shared ownership. In the podcast, Vanitha built an agent to catch anti-patterns and explain them to developers, realizing this agent should be focused on teaching rather than just reviewing code. This teaching moment can happen earlier, during the development process, rather than waiting for the commit.

The issue of cognitive debt arises when knowledge sharing is lost, leading to a lack of understanding of the system being built. Technical debt lives in the code, while cognitive debt lives in people. When teams stop collaborating, they lose track of features and their purpose, creating cognitive debt. To preserve knowledge sharing, engineers should capture their intent, architectural choices, behavior tradeoffs, and scope calls before merging changes.

By reviewing acceptance criteria instead of lengthy code diffs, reviewers focus on understanding the decisions behind the implementation. This approach keeps knowledge sharing alive, helps engineers build mental models, and prevents mental model erosion.