Ai

LLMs have reversed the traditional programming workflow - code now comes first and cheaply, while understanding follows. The locus of difficulty has shifted from construction to comprehension.

Instruction-following isn't enough; we need agents with judgment, context, and a sense of consequence.

Balancing flexibility and control is key to bringing AI agents safely and effectively into enterprise operations.

This post examines the idea that programming is less about code and more about the internal theory developers construct - a human practice of meaning-making that is not easily captured by AI of today.