AI-generated code presents a trust problem as developers must choose among multiple options without clear guidance on their tradeoffs. Traditional tests only verify correctness, leaving crucial aspects like security and performance unaddressed. SafeCode Arena aims to solve this by providing an automated verifier that scores code candidates across five simultaneous axes. These axes include correctness, security, performance, maintainability, and resource usage, each with assigned weights.The system uses a multi-axis scoring approach to quantify tradeoffs, allowing for defensible decision-making. Implementation patterns involve isolating code execution using WebAssembly for safety. Persistence of evaluation history enables tracking regressions and building an audit trail of code evolution. SafeCode Arena supports multiple languages, allowing for cross-language comparisons on the same rubric.Integration into CI/CD pipelines ensures that code candidates are scored before merging, preventing surprises in production. The core idea is that a rubric, rather than gut feeling, defines trust and responsibility. Explicit rubrics make code evaluation transparent, reproducible, and teachable.This system allows developers to confidently choose and merge code, knowing they can explain their decision based on objective scores. SafeCode Arena is already implemented with key features and plans future language support and integrations. It offers a systematic verification method to complement AI-generated code, addressing the industry's tendency to implicitly trust AI output. The project is open-source and built with Rust, Wasm, SQLite, and Python.