2x, not 10x: coding with LLMs in 2026
Summary
The article argues that 2026 productivity gains from LLMs come from reliable automated feedback loops rather than a 10x boost. It proposes a staircase hypothesis where further model improvements yield diminishing returns, and real gains come from tooling, workflows, and sandboxed environments. The author shares personal experience with iterative coding using LLMs and advises cautious, incremental adoption.