General Discussion

Vibe coding is a restatement of the goals of... COBOL. Read Jean Sammet's History and Fundamentals of Programming, published in 1969. She was one of the inventors of COBOL in 1959, the stated goal of which was to allow for "anyone" to write code by "speaking English". Of course, COBOL didn't even come close to making that goal a reality.

I used to teach Compute Science at the University of Kansas ( Go Jayhawks ), sometimes I was teaching CS200 ( Introduction to Computer Science ). On the first day of class I didn't teach a damn thing about Computers or FORTRAN or Assembler or Basic or anything. What I did was to go to the blackboard ( yeah, chalk on a blackboard - white boards and markers hadn't been invented yet )... and have the class play tic-tac-toe.

I would play the class. I would put an X someplace and then ask the class where they wanted an "O". When they all agreed on the move... I would ask a simple question "why there?" Explain why you needed to make THAT move.

It's not the code, it is solving the problem ( whether in English as in an AI prompt or in FORTRAN or C or Python ). Someone needs to describe the algorithm that solves the "puzzle". The AI can do rote recitation of known solutions, and then code it for you... but if they haven't been trained on it, good luck getting anything out that makes sense or is any way optimal.

For example, I could possibly explain the rules of Chess well enough for an AI to invent some Chess playing code... might even find enough articles about it in what it scraped from the internet to do a decent job. However, would it ever be inventive enough to invent a Chess playing program that simulated Chess moves using nothing but Boolean ( and pseudo Boolean ) logic ( AND, OR, NOT or XOR, and leading and trailing zero counts ). I doubt seriously than an AI could invent such an algorithm. However, in the early 1980s, my co-worker and I did just that.

That said, I use AI every day to help me in my work... and I use two different AI models, one to create the code, and another to verify the code. Plus I use AI to document the code better than all but the most diligent programmer.

Measuring my productivity is difficult to gage... but my organization uses it extensively now... and we are doing almost 2X the amount of work that we were doing a year ago, with only a modest increase in staff.

Asking for the ROI on AI is like asking the ROI on the Internet in 1985... it is early days of AI adoption right now... or another example... what is the ROI of smart phones? Everyone has one now... probably 20 percent of the human race either has or has access to a smart phone. What is the ROI on that? Would you give up yours?

OK, now for the full disclosure... I am a senior architect at a major AI chip maker... and have been here almost 8 years now. Doesn't mean I can't be objective about AI and some of the hype therein. Nor does it mean I am an uninterested bystander because a great deal of my compensation is in the form of stock RSUs... which I use to promote DU. DU charities, pet rescue charities, and Democratic candidates for office ( nation wide ). Without AI I doubt that I would be able to do this. Doesn't mean that AI is a good thing... just stating my obvious interest in the subject.