LogFAQs > #985925104

I've been programming for 30 years with a degree (and I started learning as a hobbyist nearly 45 years ago). Here are my thoughts:
* First, it doesn't matter if vibe coding "actually" works. All that matters is that the people selling it are good at selling it. Many of the stories we are reading about it, even those that make it sound bad or scary, are actually advertisements.
* I've vibe coded a relatively simple problem (a command-line utility that did one well-defined-but-uncommon thing). I couldn't use it as-is, but it would have worked and definitely saved me the trouble of looking up a lot of stuff. At the same time, it was only my experience and familiarity with the larger concepts (Windows UI events) that allowed me to see that it was doing the right thing.
* Most of what I develop involves solving problems that nobody has solved before. Not because they are super hard math, but because they involve unique and subtle workflows and constraints. So much time spent merely hammering out the definitions of the words we are even using to describe the problem. Coding is already much faster than analysis and requirements gathering. Vibe coding would not help me with that.
* I keep hearing about AI replacing "grunt coders". I have decades of experience and have never seen a "grunt coder". I honestly have no idea what this refers to. Wouldn't such repetitive tasks already be automated by code generators or declarative programming practices?
* I expect the next big thing to be the realization that English is an inefficient and ambiguous way to communicate with an LLM, and that we should instead use a specialized vocabulary and/or notation. At that point, congratulations, you have invented a new computer language. The history of computer language development is the story of trying to clearly tell computers exactly what we want them to do. In fact, this seems so obvious, that I believe it is being intentionally held back until the current revenue sources dry up and a new-new-thing is needed.