Alibaba open-sources AI code review tool to reduce costs

Alibaba has open-sourced Open Code Review (ocr), a tool designed to address common inefficiencies in AI-driven software engineering. The tool aims to mitigate issues found in general-purpose coding agents, such as incomplete coverage of large pull requests, position drift where defects are reported on incorrect lines, and high token costs.

By utilizing a hybrid architecture rather than treating code review as a pure text-generation task, the tool reportedly reduces token costs by 90%. This approach helps separate the cognitive burdens of bookkeeping, syntax coordination, and semantic reasoning, which often overwhelm standard large language models during the review process.

As AI-generated code becomes more prevalent, developers face increasing challenges in maintaining mental models of complex systems. Experts suggest that while AI provides significant leverage, developers must rely on their architectural experience to ensure code remains maintainable and to avoid the pitfalls of simply signing off on unverified AI outputs.