From my perspective[1], the problem with writing everything you're going to do upfront is that it's a terrible way to develop new ideas.
Outside of academia, where you just do shit to see what happens, you can get through maybe a dozen different "idea iterations" in a month, and within a short span of time you've really got an understanding of what does and doesn't work.
In academia, where everything has to be justified, it seems like it takes decades to reach the same level of understanding that you could get independently in months.
Of course, because your're more methodical and precise you can much more certain that you haven't made as many errors, but as a human being who doesnt know what the hell you're doing you'll almost always learn more from playing with something spontaneously 50 different ways than carefully planning 5 approaches.
I feel like science as a whole would be much more effective if we split it into two separate tasks: trying to wrap our heads around new things, and then only after this has done do we begin to formalise things. The current model of pretending we know how to approach new things from day one just seems to stifle progress and prevent bad paradigms from dying.
[1] I left my PhD due to complete frustration with the system after first year - kept getting pushed towards the goal of "going through the motions" and adding to the pile of polished turds that is decoding algorithms for brain-computer interfaces. Still bitter about it 6 years later, probably will be until the day I die.