The peril of laziness lost

Article Summary

Bryan Cantrill revisits Larry Wall’s classic “three virtues of a programmer” (laziness, impatience, hubris) to argue that LLM-generated code is losing the virtue of laziness — the drive to create powerful abstractions that simplify systems. He observes that vibe coders bragging about massive lines-of-code output are celebrating exactly the wrong metric, as AI tends to produce verbose, concrete implementations rather than seeking the elegant abstractions that make software composable and maintainable. The essay argues this loss of laziness poses a real peril to software quality as AI-assisted development scales.

Discussion

• Multiple commenters recalled Ken Thompson’s quote about his most productive day being throwing away 1,000 lines, and the famous Apple “negative 2,000 lines” story, reinforcing that less code is often better
• A computational fluid dynamics researcher noted that vibe-coded test suites look impressive in size but often have critical gaps compared to carefully crafted manual tests
• Several pushed back on the permanence of the problem, suggesting it may be fixable through better prompting, agent frameworks, or CI/CD steps that review code for unnecessary complexity
• The German General von Hammerstein-Equord’s classification of officers by cleverness and laziness was cited as a parallel — clever and lazy people are suited for leadership because they find efficient solutions
• Commenters debated whether judging AI output by code quality is itself the wrong frame, arguing the real measure should be value generation rather than lines of code in either direction

Discuss on HN