AI-Augmented Programming Education: Why LLMs Make Learning to Code More Important, Not Less

Abstract

It is frequently argued that Large Language Models make learning programming irrelevant. In this talk, I present the case that the opposite is true—but it requires a fundamental shift in how we use AI. Rather than using models to solve problems for us, we must configure them to teach us to become experts. With AI, we have the best teachers in our hands now, and not utilizing them to enhance our programming abilities is a huge mistake. This talk introduces a framework demonstrated through the practicego project, which shows how I set up Cursor and AI models to help me learn Go programming. The core insight is that learning to program, since the advent of GPT, is not about reading tutorials or following examples. Instead, it is about actively building programs where the AI guides us accurately to understand the concepts deeply. This explores new avenues wherein previously difficult problems become easy to tackle. The concepts that were considered advanced are now accessible and available in our engineering toolbox. In this talk, attendees will learn how to transform AI tools from code generators into personalized tutors, gaining practical techniques for setting up this teaching environment. They'll see live demonstrations of the difference between "AI solves my problem" versus "AI teaches me to solve problems." I will demonstrate how easy it is to learn Go, Rust, TypeScript, or C using this AI-first learning approach.

Reference Projects:

• practicego - founder
• Learn To Solve It - founder
• exercism - contributor
• Python - core developer