https://adventofcode.com/2024/leaderboard/day/1

I also don't see how it would be possible otherwise.

I think the strategy for the harder puzzles is to still "do" them yourself (i.e. read the challenge and understand it) but write the solution in English pseudocode and then have an LLM take it from there. Doing this has yielded perfect results (but less than perfect implementations) in several languages for me so far and I've learnt a few interesting things about how they perform and the "tells" that an LLM was involved.

  i1:("I  I";" ")0: `:1.txt;
  sum {abs last deltas x }each flip asc each i1 / answer 1
  sum {x * sum x = i1[1]}each i1[0] / answer 2

What does each flip asc do?

    from collections import *
    xys = list(map(int, open(0).read().split()))
    xs = xys[::2]
    ys = xys[1::2]
    xs.sort()
    ys.sort()
    print(sum(abs(x-y) for x,y in zip(xs,ys)))
    
    yc = Counter(ys)
    
    print(sum(((yc[x])*x for x in xs)))

    data = { i+1 : sorted([ x for x in list(map(int, open('input').read().split()))[i::2]]) for i in range(2) }
    total_distance = sum(list(map(lambda x: abs(x[0]-x[1]), zip(data[1], data[2]))))
    print("part 1:", total_distance)
    similarity_score = sum(list(map(lambda x: (x*data[2].count(x))*data[1].count(x), set(data[1]).intersection(data[2]))))
    print("part 2:", similarity_score)