AI-Assisted Code Reviews: Mario Hernandez’s Early Experiments

Mario Hernandez reflects on his early experiments with AI code review tools, treating them as helpful assistants rather than replacements for human oversight. He emphasizes that while AI can spot certain inefficiencies, it lacks full contextual understanding—making human judgment indispensable.

Hernandez reports using GitHub Copilot inside his IDE for in‑context feedback during development, and experimenting with CodeRabbit. His most satisfying experience came from Google Gemini Code Assistant: after integrating it with his GitHub repo, Gemini reviewed pull requests and offered thoughtful suggestions to improve code clarity and maintainability.

To illustrate, he describes a Drupal Twig template where he originally used chained if‑elseif logic for mapping CSS classes to media view modes. Gemini proposed a cleaner refactor using a centralized view_mode_map and a loop—making the code shorter, easier to maintain, and less error-prone.

In closing, Hernandez cautions against blind trust in AI. He argues that AI’s role is to augment a developer’s expertise, not supplant it—especially in codebases with business logic or legacy constraints. For solo devs and small teams, though, he sees real value in letting AI offer suggestions and helping rethink patterns you might miss yourself.