Ad-hoc software development

Each application works by interacting with something - calling this or that function, or, on a higher scale - opening a file, connecting over network..

So the idea is to write some minimum of code which uses all such interactions, but does not do anything except for testing them out.
Often, while writing such code, it occurs that there is a better way. Then it is easy to change the course as no important code was yet planned and written.

Once the interaction works and the rest is obvious, I quickly add the rest, knowing that all the needed stuff is ready to be used whenever needed.

This is the fastest way to build unreliable software without any planning whatsoever.