That's a curious choice of wording. I have considerable experience encoding requirements as computer programs. According to the latest Stack Overflow survey, I seem to be in the 99th percentile in terms of coding experience ;-) I find that getting the requirements right is dramatically more difficult than encoding the instructions for the computer (even if you happen to write a big ball of hairy code).

Consider that we have 3 types of errors that can occur in software: errors in requirements, errors in translating those requirements to code, and errors in expressing that code. In other words, we can build the wrong system because we got the requirements wrong, we can build the wrong system because our code doesn't implement the requirements that we gathered, and we can have errors where the implementation has effects that we didn't consider.

There is a reasonable chance that we can build systems that given requirements will encode them into working code. This can take care of the second 2 classes of errors. However, if you look at where most of your time as a developer goes to: it is rewriting the code where the requirements were insufficient or incorrect. Until we have an AI system that can replicate human levels of thinking we won't be able to go from "I need a sales report" to code because there is no way for the system to reason about what should be in a sales report -- and especially no way to ask relevant questions about the business to write a sales report that is good for that business.

No matter how good we get at automatically translating requirements to code, you are still going to need a person to gather the requirements and think in excruciating detail about exactly what you need. You may have noticed that normal people can't do that job. They will say, "I need a sales report" and you say "What should it contain?" -- the answer will be "The sales information". Keeping track of all the mind-boggling little details and exactly how it will affect the rest of the system is still going to be a job for a human -- until we can build a system that is capable as a human (and not just any human -- a human that is able to think very hard about requirements). This will not happen in my lifetime. It might not even happen in your lifetime (which by the very nature of statistics is likely to be longer than mine ;-) ).

The "hopefully!" part was meant to be a nod to the occasional situation where people build systems that are far more complicated than the problem they are solving. Hopefully you are not building such a system ;-)

I hope that cleared up what I was trying to say. I have to admit that I am not very clear on what you were trying to say.