Everyone agrees that software is to serve humans. What they disagree about is how to best make software that serves humans.
For example, lets say you write an API that translates text. Should you assume the text the human sent in is correct and return an error? That way you can make better translations when they send the right data. Or should you try to be helpful and automatically fix spelling errors you think you've found in order to make the API more human friendly? Would make it easier but reduce accuracy.
You seem to be in the second camp, reduce power of API and make them easier to use. But lots of people think that the best way to serve humans is to make more powerful and strict API's.
More than that, there may be a whole stack of software components between your API and your users. In this sense, quite a lot of software is written for computers and not other humans (or, your users are other software devs).
This is to say, "all software is written for humans" is too general to be useful - the important issues are all in the details.
For example, lets say you write an API that translates text. Should you assume the text the human sent in is correct and return an error? That way you can make better translations when they send the right data. Or should you try to be helpful and automatically fix spelling errors you think you've found in order to make the API more human friendly? Would make it easier but reduce accuracy.
You seem to be in the second camp, reduce power of API and make them easier to use. But lots of people think that the best way to serve humans is to make more powerful and strict API's.