Knowing Just Enough to be Dangerous

I recently read a post on LinkedIn—yes, I know, I shouldn’t be wasting my time scrolling through the insufferable LinkedIn feed (I was there to look up AI jobs, I promise). The post was by Andrew Ng, and I think it did a great job outlining the advantages LLMs provide to anyone with at least a baseline knowledge of [insert knowledge work skill here]—in my case, full-stack development.

He explained how someone with deep expertise in art, for example, could get exponentially more value out of image generation tools because they understand the field’s terminology, unlike a novice who lacks the lingo. And this applies across many disciplines.

I’ve noticed this firsthand when using LLMs for full-stack development. I wouldn’t call myself an excellent full-stack developer—far from it. I know there’s a lot I don’t know, and I’m well aware I’m not the best. I’d probably struggle with a single LeetCode question. But I know just enough across most areas of full-stack development to be very, very dangerous. And that’s all I need.

With that foundation, using Claude 3.7 now gives me an ungodly amount of leverage when building applications.