This article is available to everyone, non-members can view it via this link
The hype around large language models (LLMs) has made it feel like AI could swoop in, save developers from endless coding sessions, and maybe even write that dream project you’ve had in mind. Spoiler alert: it won’t. While LLMs are surprisingly good at some parts of coding, they fall flat when it comes to creating entire software systems from scratch.
It Doesn’t Get Your Vision (Unless You Spell It Out in Painful Detail)
Humans are great at turning vague ideas into working systems. You might tell a teammate, “I want a simple chat app,” and they’ll ask about real-time updates, user authentication, and whether it should handle GIFs. An LLM? It’ll give you a basic chat app that may not even compile/run, with no idea whether it meets your actual needs unless you guide it every step of the way.
Where It Works Better
LLMs thrive when you need help filling in the blanks for common patterns. Already have a half-built app but don’t want to write out that boring login flow? Perfect – ask an AI to handle the boilerplate.
