The Learning Loop and LLMs
Source: martinfowler.com
Unmesh Joshi’s article explores the enduring nature of learning in software development, challenging the notion that it can be reduced to an assembly line process. He argues that design in software emerges through implementation, making hands-on experimentation and feedback essential. While large language models (LLMs) offer powerful tools for brainstorming and setup—especially in reducing friction during initial project stages—they cannot replace the deep, iterative learning required to build robust systems. Joshi introduces a learning loop of observing, experimenting, and applying knowledge, emphasizing that true expertise comes from navigating this cycle personally. He critiques the illusion of speed offered by low-code platforms and LLMs, warning that bypassing the learning process leads to fragile systems and productivity cliffs. Ultimately, the article advocates for using LLMs as translators and accelerators, not as substitutes for understanding, and calls for a renewed focus on learning as the foundation of sustainable software development.