Developers open-source Code Review Duel for cross-reviewing code using free LLMs

Developer Zoymusk built Code Review Duel for the Flip the Switch hackathon, and the project is now open-sourced with a live demo. The experiment is built on OpenCode and lives in one Switch room populated by two free models: DeepSeek as reviewer-a and GLM as reviewer-b. Both models review the same code, check each other's output, and the human participant then picks the better review. A demo video is available at https://youtu.be/Dyu2Fs-55sg, and the repository is at https://github.com/Zoymusk/code-review-duel. The core problem the project addresses is that a single AI code reviewer can be confidently wrong, and users often cannot tell. Putting two reviewers on two different free models forces them to cross-check each other's verdicts, which makes the weaker review easier to spot. The stack is entirely free, so the setup costs nothing beyond the models themselves. So far the project has shipped as a working demo and a public repository, with the full hackathon write-up truncated before details beyond the Switch room setup and the OpenCode configuration are given. No roadmap, benchmarks, or follow-up releases beyond the initial open-source drop are mentioned in the available material.