Software development is fundamentally a learning process that cannot be automated away. While LLMs excel at reducing initial friction—generating boilerplate, translating natural language into code, and helping with setup—they cannot replace the essential learning loop of observing, experimenting, and applying knowledge. The article argues that true expertise comes from hands-on struggle and contextual understanding, not from speed of code generation. Tools like low-code platforms and LLMs offer tempting shortcuts that bypass learning, leading to a maintenance cliff when requirements deviate from standard patterns. Agile methodologies succeeded because they acknowledged this learning-centric nature of development, and the same principles apply when working with AI tools.
Sort: