MIT Mystery Hunt 2026: grab bag

25 Jan, 2026

I did the 2026 MIT Mystery Hunt with ✈✈✈ Galactic Trendsetters ✈✈✈ again. I liked my approach last year of quickly publishing more posts that are shorter and less organized, rather than putting things into a megapost that might just get stuck as a draft forever. (It didn't actually work that well because there are still things stuck as a draft forever from that year. Maybe this time.)

things I liked

Puzzles:

• Balancing Act is my favorite puzzle. It's the first puzzle I wrote down as a candidate for my favorite puzzle during hunt and nothing else contended with it. It has a fun core loop with a neat construction feature that helps you gain a foothold and error-correct, while also being relevant later in a way a teammate described as a "Chekhov's constraint".
• Keep Going! is funny. (It might be hard to postsolve unless/until the writing team ports it somehow. Also, knowing the answer is a much bigger spoiler for this puzzle than for many others, so you might not even want to open the puzzle page if your team solved it.)
• Devilish Devilries is funny in a totally different way (heed the content warning). We swarmed it shortly after unlocking it late on Friday night; it was the last puzzle I did before going to sleep.

Rounds:

• The Land of No Name was an amazing round. (It might also be very hard to postsolve without writing team support.) Solving ��������� �������������� (url 8063) with two or three given letters was a highlight of my hunt. My biggest caveat is the meta is unforgiving — we had the right idea for what to do early on, and forward-solved all but one answer, but even so, somebody tried pretty hard to backsolve the last answer and/or solve the meta and failed (I think — I'm guessing based on our spreadsheet). We forward-solved the last puzzle (��� � �� ���� (url 9350)) before solving the meta. This seems expected, based on the designer's notes.
• Atlas of Mosaics was also excellent (and also probably hard to postsolve; hmm, this might be a theme). Of the parts I did, I expect this to be an unpopular opinion based on how my team reacted, but I liked the Lifecycle "jigsaw" and wish I had gotten to do more of it — it's just such an absurd take on a jigsaw, and absurd takes on puzzles is part of why I love Mystery Hunt. (I got pulled away to work on The Tortured Programmers Department, which is certainly more in my comparative advantage, and liked doing the subpuzzles I did, but I still found the Big One too intimidating; as of time of publication, it's still unsolved. Case of the Superhero Dinner Party looks phenomenal but I think it was already fully solved when I first noticed it. The Puzzle That Cannot Be Named also seems really cool but I think there were only like four clues left when I saw it, and I only solved one or two after struggling mightily.)

Logistics:

• I thought the way the "Research Tasks" scavenger hunt tasks were integrated into the rest of the hunt was brilliant. I'd love to see future Mystery Hunts do something similar.
• The website was great, from smoothly surviving the hunt start to releasing solutions at wrap-up! I think the main thing I ran into was I would have liked to been able to click through cutscenes on my phone to look at "normal puzzles", even if large swathes of the hunt wouldn't have worked.

Moments:

• My crowning achievement was Sunday at 2am. I had returned to my room and gotten ready for bed, and my roommates were asleep, but I didn't feel sleepy yet, so I quietly joined voice chat on my phone and opened the Fate's Casino spreadsheet. Earlier, before leaving our hunting classroom, I had set up the spreadsheet so the team could collect data, which was mostly done by then, but there were a few more steps in the puzzle. I stared at the phrases, then found some words, then some more words, then somebody else backsolved our last answer, so far so good. But then we indexed the answers and they didn't spell anything. But then... I randomgrammed the answer. On that note I went to sleep. (I might quibble that the quirky wording of the flavor text suggested a more creative interpretation instead of the basically "straight" reading to me, but overall I think the thing we missed was fair — I had seen enough of the round that I should have been able to figure it out by myself.)
• Some kinds of solving nonsense are easier in person, like being part of a group of people repeating the sequence of phonemes we extracted from Squint Your Ears out loud for 20 minutes to people around them until they finally figured it out.
• I liked the event I went to, Advanced Pictograms. I'm always a huge fan of restricted communication tasks. Alas our team was one letter off from communicating all four things exactly.

have you tried?

I can't think of anything I want to revise. Halfway through a typical Mystery Hunt, Galackboard is littered with bright orange stuck puzzles that multiple waves of solvers have tried and failed to unstuck, but this hunt felt unusually smooth. I don't think we got hard stuck on any puzzle. Exactly what this means is unclear; personally I still think this is best explained by Galactic being "culturally" similar to Cardinality, in the same way we were to teammate when they wrote in 2023.

Two anecdotes:

• I spent a lot of time floating around the room or on Galackboard looking for stuck puzzles to unstick or extract from, which I think I did pretty well at this hunt. Often as a silly joke I like to start by asking "Have you tried reading the first letters?" (the assumption being that everybody would have tried that before asking for help); but it worked once this hunt (on Maps, understandably because the short cluephrase has a nonobvious word break and abbreviation). So I guess I have to keep doing it.
• The other thing I might consider adding is: have you tried just looking at it? I feel like that is how I extracted On the Fence. I don't think adding this would actually be helpful though.

misc vibes

• I felt like Galackboard's "code monkey" tag was used a lot, and spent way more time coding for way more puzzles than in past hunts. Mostly this was throwaway Noulith scripts, and for one or two of the puzzles I got to share the script with others through the URL so they could run it, so I guess my 3½-year investment is paying off! I did have to pop back into Python to call some heavyweight numpy things. (For the curious I did not use an LLM for any of this because obviously it wouldn't be good enough at Noulith. I did ask Claude for help connecting my laptop to our classroom's projector to stream kickoff, and it was moderately helpful.)
• Speaking of numpy, I did two puzzles with "hard math" this hunt, the previously linked ��������� �������������� (url 8063) and Normal Jigsaw, which is maybe twice as many "hard math" puzzles as I expect to do. Some readers may know that I was a math major at MIT and did my time in the math contest world, but there are still people on ✈✈✈ Galactic Trendsetters ✈✈✈ who know more math than I do, who were invaluable for getting these.
• Last year in my reflections I noted how I regretted almost never interacting with the radio after thinking it was really cool during kickoff. This year I spent a bit more time wandering around MonQuest than strictly necessary. It still wasn't a lot though. Early on I did an in-game Minesweeper activity even though I expected that somebody else had already done it; and once or twice, somebody asked me to help them with some interactive or other by looking over their their shoulder. I explored a little bit at the very end when we kind of ran out of puzzles to do. But I did feel like there was a bit of a learning curve to playing MonQuest effectively that, for me, made just doing puzzles a more alluring task to do during the hunt.
• The thing I felt worse about missing out on is Puzzmon: The Card Game. During the hunt, I thought it was just fodder for a specific conundrum puzzle somewhere; I didn't realize it was an actually playable card game with strategy or depth until long after the hunt. Oops!

share your no name order

JavaScript to run from your browser console on the activity log:

s="";c=x=>String.fromCharCode(64+x);Array.from(document.getElementsByTagName('td')).reverse().forEach(td=>{h=td.querySelector('a')?.href||"";if(/^Sol/.test(td.textContent)&&(r=/\d\d\d\d/.exec(h)))s+=`[${c("D{GZdeyLbOTCfzU|[BJK`^MFmP_".indexOf(c(r%97%61))).replace("@","meta")}](${h})`});s