And now, for the fun part — numbers and anecdotes!
Thanks to our puzzle command center, we were able to track the progress of every puzzler as they worked through the puzzle. Over the span of almost a month, hackers made 626 accounts and 12848 total guesses. This is an insane number, considering that the puzzle was hidden and not very publicized.
Our fastest solve time was 7.73 hours from account creation to finish! Removing suspicious completions, the average time it took our first 50 winners to solve the puzzle was 20.76 hours. Breaking down by puzzle part for all hackers:
386 puzzlers completed puzzle 1 (dogebot), with a mean solve time of 1.09 hours
299 puzzlers completed puzzle 2 (paper doge), with a mean solve time of 1.07 hours
211 puzzlers completed puzzle 3 (square doge), with a mean solve time of 3.16 hours
190 puzzlers completed puzzle 4 (barking doge), with a mean solve time of 0.57 hours
169 puzzlers completed puzzle 5 (maze doge), with a mean solve time of 1.40 hours
114 puzzlers completed puzzle 6 (stronghold), with a mean solve time of 9.38 hours
To prevent brute-forcing puzzle codes, we implemented a rate limit on guesses hackers could make. After making too many submissions too quickly, a sad doge would appear:
Based on command center logs, the codes guessed most often were related to puzzle 1, which makes sense given that most puzzlers got stuck on 1.
All this data does a great job showing us the puzzle’s reach, but I think the puzzle experience is an even more compelling story. I had lots of fun staying up late the first night the puzzle was released, watching people submit guesses and replying “meow” to any hacker who direct messaged asking for hints. Our puzzle was created not only to technically challenge hackers but also to humanize our hackathon, and I believe we made good progress towards both those goals.
As an example, a few lucky hackers were congratulated in person for finishing our puzzle. With the aid of the command center and a Slack webhook that notified us every time a puzzler finished, we saw that a couple hackers also happened to be fellow interns, resulting in some pretty funny situations.
The Slack group was a great channel of communication between puzzlers and organizers, and we had lots of fun interacting with hackers!
Although our puzzle ate up a lot of hackers’ time, hopefully they learned some new tricks. In my own puzzle-solving marathon last year, I didn’t know anything about DTMF encoding and hashing and had to teach myself as I went.
Trying to solve the puzzle is a lot like trying to build a hack — both can be frustrating, require hackers to think on their feet, and are ultimately pretty rewarding. Our team collectively spent dozens of collective hours putting together this puzzle, and we hope this experience was as memorable for hackers as it was for us. :)