Puzzle Hunts¶
I joined the MIT Mystery Hunt remotely this year as part of Spore Giant, a team with hunters primarily based in Singapore. It was a blast watching the experts on the team blasting their way through puzzles, most memorable being a puzzle we unlocked at around 2am local time with a piece detailing a debrief for a contract bridge competition. It was solved in close to 10 minutes, with about 5 hunters with deep understanding of bridge strategies and notation.
I'm still a new puzzle hunter, the CMU puzzle hunt in 2025 being my first. But fortunately my interests overlap with common puzzle enjoyers, namely board games and climbing. There was a puzzle where you have to identify board games based on blurred out covers, one on A Fake Artist Goes to New York and identifying other Oink games and puzzle that mashes up Cryptid and Cascadia, and one with Suburbia and Ark Nova. On the climbing end, there is one trivia + crossword on a kilter board and identifying sends.
Of course this is also the year that the LLMs are used extensively. Unsurprisingly most of the models were not great at puzzles, but we could use them for semantic searches and topic research. To the end if a thinking model is used, it will go into loops of thoughts that recurse.
Anecdotally Gemini worked the best for me, with this helping to unlock a crucial clue in one:
