Perhaps the best writing advice I've ever heard was to only write something you haven't read yet. Don't do stuff that's already been done. Obviously, that's easier said than done – almost everything is (the only exception I can think of is someone saying, "I think I should quit my job", but only because "I quit" is a reasonable way to achieve that). But actually doing it, easy or not, helps with the most important element of writing-as-abstract-process: actually having a thought to write.
Obviously, I've had some bad ideas here on the blog, and certainly recurring themes (although, ideally, viewed from different angles or contexts each time). But I endeavor, not just to publish text, but to commit to and share an idea with every post. Sometimes I wish to persuade, sometimes document, sometimes surprise or help someone bend a part of their mind that was previously sore from disuse.
But fundamentally, I wish to write something that hasn't yet been written, and the method I've chosen is to read a lot and write things I've not yet read. This general approach is probably the best if you constrain your writing to meaningful or interesting topics.
This might all seem obvious, but I think it's important to understand how much of communication is driven by two complimentary forces: first, to provide a minor variation on a theme, and second, to use whatever voice you have to amplify or diminish the collective voice of those around you. These are important things, I think, but in 2019 exist in such hyper-abundance I couldn't possibly make a meaningful contribution to them. There's much more marginal advantage in perspective shifts, smaller discussions, questions (as opposed to answers), guesswork, and specific imaginings of ways things might be. These are also important things, but how often do you see them?
Oh, sure, you might say, I remember that person said something wrong and embarrassing – yet making something up and being wrong is not the same as true guesswork. It might be a guess, but if you don't put in the work, you might as well just have a puppy walk on your keyboard, your cat's calculator providing the needed statistics and sums.
Being wrong is an important part of engaging richly with the world – it's essential to form fully realized guesses (and then, of course, check them). How else would we learn? And yet I don't see the desperate parenthetical warnings and explicit humility this approach requires, not as much as I'd like, at least.
Make a strange guess. Add details until it's a unique guess. Reject ones that are impractical or absurd, sure, and begin again. Push beyond your own knowledge by testing a coherent guess against the world.
It's not that unusual.