TRENDS & REFLECTIONS

What if today's best engineers aren't enough for tomorrow's systems? By following today's trends, we may discover that human-level code quality was never the problem we needed to solve.
ADAM TORNHILL
JUN 30, 2026

Having an AI produce code at the quality of a human expert has been the dream and definitive benchmark for years now. Recent progress on both foundation models and, perhaps more important, the agentic harnesses have made this dream seem less distant.

Let’s assume we (or BigTech) succeed: tomorrow’s coding agents perform at the level of today’s best software engineers. My argument is that this still wouldn’t be enough