AI stance is not binary

AI use in software engineering is often framed as a choice between handing work to agents and rejecting them to protect the craft. A relay-race analogy offers a third approach: divide tasks so people and machines contribute where each is most effective. An engineer who only watches an agent generate code risks losing practical intuition, much like a runner who stops training. But closely supervising an agent through work the engineer is doing in parallel can duplicate effort and cause burnout. Neither arrangement makes good use of the partnership.

In a relay-style workflow, engineers remain responsible for understanding problems, shaping architecture, making trade-offs, handling complex code, and judging whether results are sound. AI can help explore options, produce repetitive code, prepare tests, search documentation, or implement a clearly defined component while the engineer tackles something else. This requires engineers to keep practicing, assign work they can verify, and learn through experimentation where AI helps and where it wastes time. The measure of success should not be how many lines of code are produced, but the quality of the handoff and the resulting software. Engineering depends on knowing what should be built, why it matters, and how to recognize failure. The strongest approach is neither maximal delegation nor blanket refusal, but a deliberate division of work that preserves human judgment and skill.