Delivering code that you’ve proven works is essential in software development. The piece highlights a troubling trend where junior engineers use AI tools to submit large, untested pull requests, expecting code reviewers to catch mistakes. This approach is not only inconsiderate but also neglects a fundamental responsibility of software developers. They need to ensure their code functions correctly before involving others. Proving code functionality involves two key steps: manual testing and automated testing. Manual testing requires engineers to verify that their code works as intended, often by executing a series of commands and documenting the results. If a change is complex, creating a screen capture video can help illustrate its effectiveness. After confirming the primary functionality, engineers should test edge cases to identify potential issues. Automated testing complements this process by ensuring that changes remain functional over time. A solid test should fail if the implementation is reversed, reinforcing the importance of both testing methods. The rise of coding agents like Claude Code and Codex CLI has made automated testing more accessible. These tools can run code and check for errors, but the responsibility still lies with human developers to ensure accountability. While agents often generate tests independently, maintaining a well-organized test suite is crucial for effective collaboration. The article emphasizes that the real value in software engineering comes from providing code that’s not just written, but thoroughly validated.