Hard Parts of Agentic Engineering

Hard Parts of Agentic Engineering

Agentic engineering is the practice of building software in collaboration with AI coding agents that generate and execute code in a loop until a specified goal is met. The defining capability is code execution. Without it the LLM output is limited, but with it, agents can iterate until the software actually works. Rather than replacing software engineers, this shifts their focus toward the higher-order craft of determining what to build, specifying problems at the right level of detail, and verifying that results are robust and credible. Developers who harness coding agents effectively can become significantly more ambitious and productive, delivering more impactful software than they could alone.