Mystery Hunt 2021 [Part 2/2]

This post is the second of two entries about the 2021 MIT Mystery Hunt. You can see the hunt website with most of the puzzles and solutions at You can find my first post here.

Here, I’ll talk about some puzzles I have thoughts on. These are necessarily puzzles that I contributed to (you can find a list of those here). There will be heavy spoilers throughout!

Yew Labs

✏️✉️➡️3️⃣5️⃣1️⃣➖6️⃣6️⃣6️⃣➖6️⃣6️⃣5️⃣5️⃣: This is a great puzzle to show folks who don’t have much experience with puzzlehunts. We’ve put up a replacement to the texting interface, so you can still do it from the site itself! Lots of teams got mad at us because the answer was literally EMOJI, and they felt like they could have just guessed this from the start and skipped the whole puzzle. I found this reaction kinda strange given that no team actually managed to do this (as far as I know). Also, this is only vaguely related to this puzzle, but if you haven’t seen it before check out the greatest fruit math problem of all time.

Hey, Can You Give Me A Hand With This Puzzle: This is one of my favorite puzzles in the hunt; the extraction step is nothing short of brilliant. During the hunt we realized that Google Sheets couldn’t handle the load of so many people viewing the spreadsheet at once (even though it was in view-only mode), and we had to make extra copies on-the-fly.

Just Index: One of two puzzles in this hunt that pokes fun at the concept of indexing in puzzlehunts.


: As CJ notes in his post of galactic proportions, no testsolvers actually solved the Miracle Sudoku since an earlier version of this puzzle made it too easy to solve using Nutrimatic. Thanks to Anderson who suggested using the red symbols to add an ordering step to the puzzle to prevent this.

I highly recommend this Planet Money episode about the Chicago Boys and the Miracle of Chile. Also if you haven’t seen Nolan’s film Dunkirk (about the Miracle of Dunkirk) or Sully (about the Miracle on the Hudson), those are both great as well. (I’ve never seen Miracle (about the Miracle on Ice).)

Alternate Controls: One lesson I learned from this hunt is that if you must have a cyclic extraction step involving a short answer or cluephrase, it’s super useful to provide an enumeration to solvers to help with the added ambiguity about where the starting letter is or which direction the cycle goes. We did that here to good effect I think.

Circles: Like Lewis, I have a fascination with Mystery Hunt round structures, so this was one of my favorite puzzles in the hunt. The jigsaw pieces were originally just dots which made the aha a lot harder.

Unraveling the Mystery: What a cute idea for a puzzle! I liked it so much that I let the fact that we now had two different Student puzzles about past Mystery Hunts slide. We were (very mildly) worried that Randall Munroe would publish a comic on January 14th 2021 which might break this puzzle.

Baker House: This meta proved to be super hard during testsolving and despite our best efforts I think it still ended up being the hardest Dorm Row meta by a fair bit. For whatever reason people had a lot of trouble interpreting the yellow arrow pointing down. (It also didn’t help that many people are unaware that the fifth solfege note is SOL and not SO.)

MacGregor House: Someone asked me why this meta required three authors. My response:

One person said we need to write a giraffe puzzle meta, one suggested we need to make that meta MacGregor since it “kinda looks like a giraffe” and one person picked the actual answers so that it worked for Dorm Row.

(I’m person #3).

Simmons Hall: This is one of the coolest metas in this hunt IMO. It was noticed by someone that two different Student metas are each resemblant of a Derek Kisman puzzle. (Simmons Hall to In the Details and Random Hall to 50/50).

Green Building

Voltage-Controlled: What a great puzzle. One of the moments that stands out the most to me when testsolving it was when I recognized that one of the melodies was Axel F but didn’t know the name and had to repeatedly hum it in our Discord voice chat until someone joined who could identify it. I still think the last step of this puzzle is a tad too hard, and wish more teams could have solved it.

Green Building: As Nathan notes in the Authors’ Notes, this was the first puzzle written for this hunt. Many teams (especially those with MIT students/alumni) were able to solve this without even needing the intermediate cluephrase. We were aware that this was possible and decided that it was fine (and maybe even desirable) to give teams with MIT students an advantage.

Infinite Corridor

The idea to make this meta have an infinite number of puzzles was present long before we had any idea of how to actually write such a puzzle. I’m really happy with the resulting puzzle Jon came up with, and I’m sad I didn’t get to experience solving this round in an actual Mystery Hunt.


As I noted in the wrap-up, The Athletics round was the last one that we finalized. In particular, we were starting to be worried that our hunt was too long, so we intentionally waited on assigning answers for this round until we were sure that (1) we had the manpower to write an additional 25 puzzles and (2) we were sufficiently sure that teams could handle an additional 25 puzzles.

Holding off on assigning answers also allowed us to repurpose subrounds to address various issues with our hunt. For example, we had long wanted to have a bunch of puzzles in our hunt that used the MMO, but because the MMO was still in the process of being developed, these puzzles were largely blocked. It was useful to be able to make the entire Football round a bunch of what were essentially MMO mini-games that weren’t even required to have a particular answer. These field goals were received pretty well; huge thanks to Herman who spearheaded the effort to make them.

Similarly, it was nice to have a round of puzzles that time-unlocked—as Baseball did—because it allowed us to ensure that the scavenger hunt, runaround, and baking puzzle all unlocked during the day instead of the middle of the night. It also let us announce when the piggy bank puzzle would unlock so that its owner could make sure they were free at that time.

Time-unlocking the Basketball round (which had our teamwork-time puzzles) ended up being a poor decision. Our reasoning was that since these puzzles were some of the most dynamic and fun ones, we wanted all teams to see them on Saturday, regardless of their progress through the hunt. In practice every time we unlocked a new Basketball puzzle, the load kind of overwhelmed the server hosting them, rendering all of them unplayable. Lesson learned.

Squee Squee: This puzzle was inspired by this comment I made all the way back in January 2020 in our #bad-ideas Discord channel:


Huge thanks to Chris for actually writing it and making it happen. Many, many teams were unwilling to smash their porcine friend, and instead attempted to remove the paper inside or backsolve the puzzle. I fielded a phone call from one team who informed me that their five-year-old was particularly attached to the piggy and really, really didn’t want to break it. (I encouraged them to break it anyways, but apparently they had already extracted the paper by other means and this was totally unnecessary. I’m so sorry!).

So You Think You Can Count: Testsolving this puzzle was maybe the single greatest hour I spent on hunt all year. By the way, did you know that there’s an easter egg when you get to 420:

blaze it


Clearly we overshot the difficulty of this round by quite a bit. Only ten teams managed to make it past ⊥IW.giga before Palindrome found the coin and we auto-unlocked the whole round, and this was even after a significant nerf we made to Rule of Three. I think there were a number of causes. The first is that it was difficult to get an accurate sense of how this round would play out during testsolving. When we testsolved the meta individually, we had to explicitly designate the puzzle Twins as being special:

if you try to open this puzzle, you get something that is too small to see

Obviously, reading this text is a very different experience from actually seeing a super small image as a puzzle. We did get one shot at a realistic hunt condition during one of our big hunt testsolves, but unfortunately, that testsolving group actually failed even earlier and didn’t even understand that Rule of Three was a metapuzzle; we had to intervene and give them a hint, which kind of ruined our one chance at getting a clean testsolve. In response, we made it visually very clear on the round page and all puzzles page that Rule of Three was the meta, but we never got another chance at re-testsolving the round.

In retrospect, I think we should have given up on being cute, and signposted what to do a little better. Maybe we could have included some flavortext on Twins that says “Hmm, this puzzle seems too small to solve…”

I think another cause of our issues is that Galactic is naturally more guess-happy as a team. We certainly didn’t envision that some teams would have a no backsolving rule in place, and I think we generally hoped that even if teams weren’t sure what to do to make progress, they’d attempt to backsolve Twins anyways, because what else could they do?

Another issue was that the puzzle A Routine Matter was too hard. We recognized early on that this answer was critical to understanding how to forward-solve Rule of Three (it’s involved in every syzygy that Twins is not involved in), and we made a note in Puzzlord to “make this puzzle relatively easy.” The final version of the puzzle was at least a lot easier than the initial version was, but we clearly didn’t go far enough. I think any puzzle that involves carefully reading through a 162-page PDF was going to end up too hard for this answer, and I wish I’d intervened a little earlier.

Finally, this is unrelated to this round but I didn’t know where else to put this: did you know that we specifically modified the perpendicular symbol glyph in the Metropolis typeface so that it looked like an upside down T? That’s why it looks so good on our site and so bad on this blogpost.

100000…0000001000000000000116: Some puzzle authors like Derek Kisman have a unique style that makes any puzzle they write recognizably one of theirs. I think Lewis’ puzzles also have this property. In many ways they remind me of Australian-style puzzles which tend to front-load a big aha that is difficult to stumble across, but feels exactly right once you do reach it. This puzzle—which we took to calling “one bajillion one gazillion one hundred and sixteen” verbally—is a great example. There’s not much you can do without the initial aha, but it’s very fair and carefully hinted at by the flavortext. I got a chance to pre-testsolve this puzzle, and it ended up being one of my favorites.

Twins: Did you know there was an Earth-Jupiter-Saturn syzygy about a month before Mystery Hunt? It extracts to O in case you were curious 😛. Incidentally, someone remarked that technically we should call Rule of Three, Twins, and Level One mesapuzzles instead of metapuzzles because of how they work. I thought this was too cute not to share.

Circular Reasoning: This puzzle is as astonishing construction. If you know someone who likes cryptics you should absolutely show them this.

Water Bottle: Puzzles involving taste identification are really something you can only do in a large event like Mystery Hunt, so I’m glad we got a chance to develop this puzzle. I bought a bunch of hint waters to help decide which five flavors we’d use, and I learned that there’s a good deal of variation in how strong the flavor profiles are. From what I recall:

How to Run a Puzzlehunt: This puzzle sorta serves as our advertisement for gph-site, an open-source repository for running puzzlehunts that we published based on the code we use to run the Galactic Puzzle Hunt. I encourage you to check it out if you’re considering running a puzzlehunt of your own! If you don’t think you’re technically savvy enough to build your own website, you can explore some other options in the “Hosting” section of Suggestions for Running a Puzzlehunt. I updated this guide in response to some things we learned from running Mystery Hunt.

The other big community contribution we made is Puzzlord, a web application for managing the puzzle-writing process that served as our replacement to Puzzletron. I wrote a little bit about Puzzlord in the guide and you can read much more about it in Brian’s blogpost.

Telephone: The intention with this puzzle was to make solvers think that they were getting real Slack / Discord / whatever notifications while solving it. This seems to have worked from what I heard.

Nutraumatic: Jakob and I started this just knowing that we wanted to write something that emulated Nutrimatic’s interface but was actually a puzzle. Extending the uppercase operators beyond A, C, and V was my idea and we worked together to come up with the particular operators and the IXPOWHXQELZTBUMXDS string, but all of the implementation work was done by Jakob. The goal was to showcase what kinds of things you could do using Nutrimatic’s engine, and in that sense I think we succeeded. We cheated on some of the operators though. For example, X works just by combining the results of the 26 queries where you assume each X is a different lowercase letter.

⊥IW.nano: I’m pretty happy with how this puzzle turned out. The final answer, the subject matter and the solve path are all super thematic with the round as a whole and I think it made for a very compelling solve. My hope is that as solvers were zooming in to the spot where the missing alum should be they would say to themselves “this is crazy, there’s no way this will work,” and then they’d spot Feynman and go 🤯.

Stata Center

PClueRS: I think this is the hunt’s best word puzzle. It’s a super impressive construction and a lively solving experience. I had a great time testsolving this puzzle.

Stata Center: We must have testsolved and revised this meta at least seven times. I think the end result is pretty great. The folding aha is clued just enough to remain fair but still be a fun realization when solvers think of it. Also, I stayed up until 3am one night in order to get the interactive folding CSS in the solution working properly so do make sure you look at that.


Don’t Get Eaten!: The reason I’m listed in the Art credits is that I actually created the image of ⊥IW’s Alchemist Sphinx. During one of our virtual writing retreats, Lennart gave a talk about how to use Inkscape to create vector graphics for puzzles, so that was the tool I used. I first put together an Alchemist-like texture by writing out a bunch of mathematical symbols in Times New Roman and rotating and attaching them to each other. Then I took an image of an existing Sphinx and masked it against the math symbol texture. FInally, I applied shading based on the coloration of the original sphinx image to bring out the features and voila:


I’m positive there were many people on the team who could have done a better job than this, but our artists were very busy with the actually important buildings on ⊥IW, so my version was what we went with.

Enclosures: Very cool puzzle; I’m sad I didn’t get a chance to testsolve it. I did factcheck it however, and I learned so much about calendars while doing so.

Clusters: This subject matter is such perfect puzzle fodder, I’m surprised I’ve never seen it used before. We thought this would be one of our easier metas but it ended up being fairly tricky. An earlier version used the final cluephrase BEYOND / TONGUE instead of ULTRA / GLOSSA, which meant that the Greek puzzle title’s starting letters were A, B, G, D, E, Z instead of just A, B, G, D, E. The lack of the Z made the ordering significantly harder I think. Also, there was an unintentional red herring with the round art (the building map is gridded) that may have led a few teams astray.


Bingo: This was another puzzle that was difficult to testsolve under real hunt conditions. We tried our best during one of our full hunt testsolves, but since it was unclear to us whether our testsolvers had more or less knowledge about our puzzles compared to a team which had just done the hunt, it was hard to draw firm conclusions. I think the one thing we overlooked is that this puzzle is much harder if you got it unlocked via time-unlocks instead of by legitimately solving enough puzzles to unlock the Tunnels round (because, presumably you’ve solved way fewer puzzles in the former case, and can thus make fewer inferences).

Disorder: I was part of a group that successfully testsolved this puzzle and part of a team that successfully solved The Seven Empires from Teammate Hunt, but I still only have the vaguest idea of how the board game in question works. Let me suggest that if the Diplomacy community wants to see even more Diplomacy puzzles, they really need to standardize on a set of territory codes to use.

Ignorance: What a monumental achievement of a puzzle. My brain broke multiple times when I testsolved it. The authors introduce a genre of puzzle—epistemic riddles—and then proceed to stretch this genre to its utter limits. At the end of this puzzle you’ll genuinely feel like you have learned a new area of mathematics. Suggested soundtrack: I Know (A Song in Ten Words).


We didn’t really start writing the final runaround until the middle of December. This was partly because we were super busy with other stuff beforehand, partly because we still didn’t have a good sense of how long our hunt was (and how short our endgame should be to compensate), and partly because we still weren’t completely sure what level of access we’d have to MIT’s campus. Given that it was all done at kind of the last minute, I’m very pleased with how it ended up.

The idea that since the universes are perpendicular you need to use the Pythagorean theorem before you can perform triangulation was both super thematic and super clever. If teams could have actually been on campus it would have been cool if they were able to walk directly to the results of their triangulation themselves, but hopefully they enjoyed watching our “UROPs” do so on their behalf. (Huge thanks to Mark, Yannick, and CJ who braved cold temperatures, rough winds, and snow in order to livestream the endgame so many times).

We had the idea that the coin would be a regular quarter that Professor Yew had left in a vending machine in ⊥IW pretty early on, and we ended up choosing the final coin location because it was a central area on campus with access to a bunch of different vending machines.