How Our Engineering Team Uses AI, Part II: Meat Proxies

Eight months after their initial report on AI in software development, the engineering team at MetalBear shares how AI has shifted from a peripheral tool to a central part of their workflow. Most engineers now rely on AI agents to generate code, fundamentally changing the developer's role from primary author to reviewer. While this shift has increased productivity, it has also introduced significant challenges, including mental fatigue from constant code review and the risk of becoming mere 'meat proxies' for AI output. The team notes that while agents excel at research, scripting, and boilerplate tasks, they still struggle with high-quality Rust code and often require human intervention to maintain architectural integrity. To manage this, the team has implemented shared context repositories and rigorous review processes. Ultimately, the transition highlights a growing need for developers to maintain deep system understanding while managing the complexities of AI-assisted development.