Advent of Code analysis through the years

Advent of Code analysis through the years(jvanelteren.github.io)

20 points by chkas 1 year ago | 4 comments

mckirk 1 year ago |

It would be very interesting to have solve times for more than the top 100 spots this year. If you analyzed those, I think there's a decent chance you would see a clear bimodal distribution -- one for people coding 'legit', and one for the LLM cheaters. The prevalence of cheating this year really was quite something: people didn't even try to hide it, and proudly occupied the highest leaderboard positions with impossible solve times, using their real names and GitHub profiles.

I can't help but feel that with this, the 'golden age' of AoC is over, because the overlap between coding puzzles that LLMs can solve in seconds and puzzles that humans enjoy solving in their free time during advent has become too large. One suggestion on reddit, where people discussed this issue, was to get rid of the global leaderboard entirely, and only have smaller private leaderboards, maybe even with moderation that could deal with the blatant cheaters. But that would also be kind of sad, since this idea of global competition definitely was part of the AoC charm, at least for the competitively motivated people. We'll see where it goes next year.

For now though: Thanks Eric, if you read this, for the years of joy AoC has brought (and hopefully will continue to bring) to so many people :)

sontek 1 year ago | |

I think it ruins the people fighting for the leaderboard spots but thats a small percentage of people. Everyone else just does it for fun or creates custom leaderboards with their friends and it'll always be fun

y42 1 year ago | |

how do you define "golden age"? personally, I still love those kind of riddles, I don't care who's cheating or not. It's like sudoku. and I bet I am not alone.

mckirk 1 year ago | | |

There definitely are quite a few different ways to participate in AoC, and for the people that are smart enough to just see it as a great opportunity to solve some fun coding puzzles and maybe try a new language, nothing much has changed or will change because of LLMs.

But for those of us who are wired a bit more simplistically, who get up at 5:40am to at least be in a semi-coherent state of mind when the puzzles unlock at 6am because they can't resist the siren call of shiny but meaningless points on a virtual leaderboard; for us it is a bit disheartening to see people solve the challenges in the time it takes us to scroll through the challenge text. And that's what I mean with the 'golden age' being over. It's not a matter anymore of 'well, if I practice competitive coding more and finally get proficient with vim, I might manage to land among the top 100 next year', because next year the official leaderboard might very well be dominated exclusively by people using LLMs.