Learning Programming in an Age of LLMs
Points and comments are a snapshot, not live.
A developer who built a system with AI now struggles because he doesn't understand it deeply enough.
The author responds to a reader who built a large TypeScript system using LLMs despite no CS background, but now finds the project above his level of understanding. The author, a veteran programmer, admits uncertainty about AI's impact but leans toward disliking it, fearing mass knowledge-worker unemployment. He argues human learning speed is the bottleneck, not teaching tools, and prefers asking LLMs only falsifiable questions. He doubts AI lets people build understanding as fast as they can build products, and suggests learning a trade like carpentry as an alternative path.
What commenters are saying
Commenters largely agree that LLMs let beginners build beyond their understanding, echoing the article's core concern. The author of Python Crash Course chimes in, saying people can learn faster with AI but must intentionally build understanding alongside projects. A commenter notes the parallel to hiring a Fiverr contractor: outsourcing the building doesn't teach you programming. Another argues that learning requires failing on your own, comparing AI coding tools to machine translation for language learning. Several suggest vanilla web technologies (HTML, CSS, JS) as a better learning path than frameworks.