Thank you to Eric Wastl for creating Advent of Code.
My method of solving the Advent of Code problems is to read the problem description, think of a straight-forward way to represent the problem input as a data structure, which is influenced by my initial thoughts about how to code an algorithm to obtain the required answer. Then write the code to first read the input file into the data structure and then implement the algorithm to get the answer. I try this on the worked example given in the problem description, then on the real puzzle input. I hesitate for a short while, looking at the code to see if I've made some silly mistake that was not revealed by the code working correctly for the given worked example. Finally, I submit my answer and hope to get a gold star like I'm still in primary school (a fairly consistent feeling for me). Once I have the correct answer I move on to solving part two in a similar way. Sometimes the data structure I chose doesn't work so well for part two, so it has to be changed and part one modified.
Once I have the correct answer to both parts I look at the code again to see if there are some simple optimisations that will improve the performance. Often there are. Sometimes something strikes me and I realise there is a much simpler way to solve it.
Sometimes the hack that produced an answer for part 1 is too inefficient to work for part 2 and I'm forced to think again. It's enjoyable to create something that works efficiently and looks elegant. I don't always achieve this, and sometimes I don't try.
I remember reading one AoC puzzle that said something like "it's 25 milliseconds to midnight on December 24 so just enough time to solve all 25 puzzles." So I have it in my head that any solution that takes more than a millisecond is doing it the wrong way. Often my solutions are nowhere near 1 millisecond. Eventually I've had enough and publish it to GitHub. Sometimes at this point I look at Will Killian's solution to the problem: it will invariably be very different to mine and much faster. This is very good for me because it means I learn something new. It also makes it a little less fun because I don't look forward to seeing how far off the mark I might be. So sometimes I don't look at anyone else's solutions until AoC is done and dusted for the year.