What it is — An essay arguing that LLM-assisted coding boosts output while quietly eroding developer satisfaction, as the job shifts from writing code to endlessly supervising and reviewing it.

Key points

• The old reward loop (solve a problem, watch it compile) is gone; developers now mostly review machine-generated code, which is less satisfying even when it's faster.
• Running several AI sessions at once tempts an endless "one more iteration," stretching work hours without a matching gain in satisfaction.
• Draws a parallel to the responsive-design shift: the skill isn't disappearing, it's moving toward architectural judgment and taste for quality-gating higher volumes of output.