The Human-in-the-Loop Is Tired
4 hours ago
- #LLM programming
- #developer experience
- #AI impact
- LLMs make programming genuinely useful but also destabilizing, leading to burnout if ignored.
- Development with LLMs introduces 'fatigue of supervision': reviewing AI-generated code lacks the satisfaction of traditional programming and feels lonely.
- AI increases work intensity and parallel task initiation but doesn't improve thoughtful completion, creating a 'human reward function problem'.
- LLM-assisted programming shifts skills rather than eliminates them, emphasizing taste, nuance, and architectural expertise as key differentiators.
- New skills are emerging, like pre-mortem analysis and encoding expertise into LLM instructions, showing evolution rather than death of craft.