I literally made this comment in a prompt in a recent ChatGPT interaction: "You have got to be kidding me, I have a rule that stated that I never wanted you to do that, and you just ignored it?" And I think the core point of this article - that vibe-coding using an AI isn't the cure-all it's made out to be - is a good one. Part of this is because AI can be stubbornly mule-headed. But just as often, it's because the problem or project isn't as clearly formulated in your mind as you thought it was. The unexpected use case, the one-in-a-million exception, the forgotten data type: humans often forget them, and AI will definitely forget them. We need to keep this in mind because a lot of these errors are the sort of thing that will impact only one in a hundred or one in a thousand people - and in institutions, we're not really good at listening to so small a demographic.
Today: Total: [Direct link] [Share]
