/dev/joe's Experience at MIT Mystery Hunt 2026

For the 27th year in a row I participated in the MIT Mystery Hunt, and this is my story.

Before Hunt

For the second year in a row, I hunted with The Team That is Now Named Later. My previous team, The Team Formerly Known As ... The Team to Be Named Later, disbanded since the 2025 Hunt and a number of the members joined Now Named, so we were significantly larger this year. In addition, a few people from TTFKA... joined Providence, and the huge mass of Caltech students apparently joined some team with mostly MIT students. And that was as much of the team shuffling as I was paying attention to.

I did a number of pre-hunt puzzles, not all necessarily as practice. I did the official pre-hunt MIT Mystery Hatch, which introduced us to the idea that there was a conference dedicated to cryptozoology (if you aren't familiar with this term, it refers to creatures that are probably mythical but some people believe are real, like Bigfoot and the Loch Ness Monster). We, on behalf of a group known only as The M.I.T., were crashing the conference by submitting a paper and claiming to be true believers.

I did Ucaoimhu's usual cryptic. I did the issue of Panda Magazine that came out the weekend before Hunt. Black Letter Labs sent a small puzzle set to advertise their return to selling puzzles baked into convincing impossible anachronistic artifacts, which I solved. At the time I post this, the link will show the puzzles, but it's just their home page, so it will eventually change to something else. I did a couple other puzzles privately shared among my teammates.

On the morning of hunt, while I was loading my car with the numerous supplies I personally bring to Hunt for a combination of my personal use and my team's use, Cardinality (who ran the hunt this year) sent out this link to the livestream of kickoff.

Kickoff

This year's kickoff was a presentation from CRYPTIC, the Cryptid Researcher & Young Paranormalist Trainee International Conference. We were introduced to this theme via the pre-hunt linked above, which concluded with teams being asked to submit an abstract for a paper to be presented at the conference. The titles and team names from these submissions ran during the countdown before the actual skit.

Victoria was trying to open the conference, when Burnham entered carrying the cryptid egg that was to be the highlight of the show. He was trying to save the egg from the MIT (not that MIT; the group we worked for in the pre-hunt, which appeared as two guys in suits calling themselves the Monster Investigation Team), a secret government organization who is trying to take the egg.

A struggle over the egg ensued, and the egg flew across the room, hit the ground, and hatched. The being that came out looked like an egg wrapped in a white tea cosy, with a crown on the side of the egg. It was immediately named The Child.

Then a portal opened over their heads, and whoever is beyond the portal sent Morse code spelling COME HOME. The people from CRYPTIC recognized that the beings beyond the portal communicate using puzzles, while the MIT were clueless about puzzles. Also, they discovered that these are Puzzmon, like Pokemon but with puzzles. So we were going to be exploring the Kingdom of the Puzzmon, which is divided into different zones, which were rounds in the first part of the hunt.

They gave us a link to the web site, and there was one puzzle already there, which contained a video explaining how this hunt works. And it needed a video because the structure was a little unusual. There are both puzzles and research tasks, and watching this video was actually a research task. Completing a research task earns you research points (RP). The video was an unusual research task because just watching it got you credit. Most of them were like the scavenger hunt tasks, do something or collect something and submit evidence that must be judged before you get credit. There were also a number of events, which, rather than being run all at once, were run in multiple sessions and you had to sign up for a time slot.

Most puzzles must be unlocked by spending 1 RP, or 2 for a meta, or 3 to open a new zone. In addition to this, we are limited to having no more than 8 of these puzzles unlocked with RP open at one time, though we were told at the start that this would change later. It turned out to be just that they opened it to 10 and then 12 for everybody later in the hunt, when it didn't even matter for us, and 100 at the very end after the coin was found.

On top of this, there is a little video game they called MonQuest that provides a world you can walk around in, similar to some of the Pokemon video games, with intentionally blocky graphics. In order to actually access most of the puzzles in the Kingdom of the Puzzmon (before you can even see them to unlock with RP) and some of the tasks, you have to explore the world inside the video game, talk to the people there, and do stuff. I did explore the game a little, but I did not do much in it because we had plenty of other people doing most of those things as soon as we got access to them.

For that matter, I only completed one research task the entire hunt, very early on, one that wanted a screen shot of a Google maps image of a building shaped like a letter. This was trivial because they hinted in the title of the task the building they were looking for.

Another part of the site, accessible from most anywhere via an icon at the bottom left of the page, is MonArch, which is like the PokeDex of Pokemon games. Each solved puzzle from the zones (but not from the dimensions we unlocked later) let us befriend one Puzzmon, and each one was added here with a bunch of information about it that got used later in some puzzles.

Start of Hunt

The zones opened up at various points, the first one at 1:05 PM when the hunt was scheduled to begin, and the rest (as explained at wrapup) based on achieving a certain level of RP.

I didn't work on any one puzzle extensively at the start because we had a big team and a limit of 8 open puzzles, so every puzzle had a lot of parallel work happening. Exsportise was our team's first puzzle solved at 1:17, just 12 minutes into the Hunt, and it happened before I ever looked at that puzzle. I did my one research task at this point (it was the third one we completed). I looked at Replacements only to find that they were just finishing the last KenKen that needed to be solved, and they got the answer just as I was understanding where they were.

At 1:25 we opened Bubble Cove, and at 1:30 we unlocked the puzzle Mechanical Soft Diet, the first puzzle I contributed in any real way to. I applied some quick Python to locate where the letters were in the long number.

I participated in the solving for Crossloop, a sort-of crossword puzzle. It went pretty quickly, but at least I contributed to that. I also helped solve the cryptic in Central Precinct and Lockup.

I spent a while on Maps, a Geoguessr type of puzzle, and pinpointed some of the images. And I solved some of the clues in Cat-astrophe.

Land of No Name (and some other puzzles)

At 2:54 we opened our first dimension, Land of No Name. Notably, all the puzzles in this dimension unlocked along with the dimension itself, so instead of 8 open puzzles at once, we suddenly had 26 more of them here. Unfortunately, the gimmick in this round was that all the letters in the puzzles were replaced with the same generic symbol. However, the puzzles were of greatly varying types, so this made some of the puzzles completely unsolvable and had almost no effect on others. I and a bunch of others started focusing our efforts mainly on the puzzles in this round with few letters to be obscured.

The first puzzle we solved in this round was a puzzle whose title was a frying pan emoji followed by two words. It's a bit awkward to even write about this round because the titles themselves could be spoilers relative to how we experienced the puzzles, and the titles are all revealed in full on the web site now (at least for me, since we solved everything). So I'm going to refer to them by the number in each puzzle's URL, which was also how we referred to them among ourselves while most of the letters were hidden. This one was puzzle 1906.

It would have been helpful to have those words, but we solved it anyway. There were some color paint by numbers puzzles. I was working on one and other people worked on others, but someone put them into a solver and contrived some way to paste the results into the sheet, making our effort unnecessary. Below these puzzles, there was a huge grid of emoji.

The PBNs solved to more emojis.

Other people seemed to know what to do with the emojis, so I went looking at other puzzles.

The gimmick in this round was further revealed when we solved puzzle 1906 at 4:13. The round page has an image reminiscent of an Aztec sun stone, with 26 little boxes around the edge which are links to the puzzles. The box for this puzzle was filled in with the letter Q, and all the Qs in the titles and text of the puzzles in the round were now revealed. There were surprisingly many Qs in some of the puzzles.

Clearly this was all set up this way. This puzzle with letters only in two words in its title was intended as part of the bootstrapping for the round, something we could solve with the letters hidden, and it provided little help for the puzzles that consisted mainly of textual clues, but it did help with other puzzles.

2871 was the second puzzle we solved in the round, at 5:03. Its title was the only completely unobscured one, being three emojis, and the text consisted of various emojis, numbers, set theory symbols, and some other symbols, with almost no letters. I naturally looked at this one, but other solvers had identified it as being related to a specific video game I didn't know (unsurprising given that one of the emojis in the title is a video game controller), and we had people who knew it working on it, so I didn't spend much time here. This one revealed the Ms for us.

I actually contributed to the puzzle whose title was a list of three words with commas, but the second word was replaced by three question marks instead of the usual obfuscation, puzzle 8402. This one has a few emojis, some equal and unequal signs, a few enumerations, and a whole bunch of pictures of board game boxes which had the names obscured. I and several other people started by identifying as many of the games as possible (which are repeated in some of the grids). Google Images could identify some of the games directly, while for others it offered nothing useful or misidentified them as similar images, but we got them with our own knowledge of the games and other ways.

We figured out the enumerations at the bottom were an alphabetized list of all the games pictured in the puzzle. The second step was in figuring out what to do with the emojis at the top. They were an airplane, a framed picture, a six-sided die, and a name badge. Someone noticed that all the games involved either planes, ships, or trains. So could the plane emoji be related to this? The four equal and/or unequal signs given with the first three sets of 12 games, and the fact that there were 12 games in each group, led someone to hypothesize that the title was Game, Set, Match with the word Set as question marks to emphasize it. We were playing the game Set!

This mathematically inspired game can easily be modified to play with any other set of objects with four attributes that can have three values each, and one of the attributes here had just been determined to be whether the game involves planes, trains, or ships.

Before long, we figured out the picture frame indicated the shape of the box: vertical, horizontal, or square. The fourth meant how many words were in the title. It took us longer to figure out the dice; we had none, one, multiple before we corrected it to none, standard, custom. And of course, the equal and unequal signs indicated whether the set in those groups of games had all the same value for the corresponding attribute among the four emojis, or all different. Finally, the enumerations didn't match the games, but we could extract letters from the names of the games in the sets using these as indices to start spelling out a message.

I wrote some code to help with the identification of the sets, getting two sets for one group and none for another, showing that there was an error. Pretty soon we guessed the correct attributes for the dice.

There was one last bit. The message was incomplete, needing the name of a game to complete it. This game was identified at the end of the puzzle with a group consisting of only three games, indicating that we needed one that completed a set. This was done quickly and we solved the clue.

Some people sitting near me were working on puzzle 2129 from the beginning of the round. This one mainly of pictures of Mah Jongg tiles, and since I don't play the game, I wasn't looking at it. This one gave the Zs.

With four letters filled in, we could now understand another part of the gimmick for this puzzle. The letters were in alphabetical order around the sun stone, and the list of puzzles at the bottom of the page was in the same order as the ones in the sun stone, in the order of the alphabet. This let us know which puzzles to look at if we wanted to unlock specific letters, but useful ones like E had puzzles which consisted mainly of words, and so were not solvable until getting others.

I did work on some non-Land of No Name puzzles during this time period in zones we were opening without my direct help, including Uneven, a clued word search in Elder Drifts we unlocked at 6:00 and solved at 6:26, and Strand-Type Game, which we unlocked at 7:11 and finished at 8:18, with several minipuzzles each similar to the New York Times' Strand puzzle, but...

with leftover letters. Once someone realized the leftover letters could be used to link the individual minis together, it was finished quickly.

the numbers are all prime.

The second word of the title, ????O??Z???O??, was FACTORIZATIONS, and I also correctly guessed most of the flavor text.

While I was looking at other puzzles, somebody else had figured out something I might have done myself, to note that

primes that are 1 mod 4, despite not having integer factorizations, can be factored into complex numbers based on writing the number as the sum of two squares. They wrote out the roots of these squares in the spreadsheet. In between the previous two puzzles I mentioned, I returned here and saw this, and noted that we had a 5 or 6 digit number and a 3 or 4 digit number in each pair, but we didn't know what to do.

When I finished the second above puzzle, I returned here and saw that somebody had figured out what to do, and by the time I caught up, it was solved at 8:29. This was the G.

Meanwhile, at 8:31 my teammates solved 3842. This puzzle consisted of six seemingly identical, redacted sentences at the start, but several letters had been filled in, differently in different instances. Also, there were a lot of Qs and Zs. One of my teammates figured out what to do, and this gave the A.

Once I got us started on 7642, the puzzle fell quickly. This was another puzzle without any letters, except for some likely to be x in identifying the dimensions of a rectangle. In early attempts, I noted...

that there were only nine different line lengths, and 9 lines worth of numbers at the bottom, and 9 different characters that made up the whole set of lines. I also noted how the shortest ones are only made of two different characters, save for a single plus sign. Someone else left a comment suggesting wrapping the lines at some position, but hadn't done it, maybe because the dimensions did not match.

So I came and did it. Specifically, I looked at the middle three lines of the shortest length, and chose a wrap width that made the groups of equal signs match up well. Soon I had all five groups wrapped this way, and discovered that by arranging them a certain way, they made a crude ASCII art map of Alaska, but one pieces of the map was missing. I started work on another set, making a soccer ball in the same way.

At this point, somebody noticed that the dimensions of the combined images matched some of the numbers at the bottom of the puzzle, with the number of pieces that the puzzle should have had in parentheses. Before long somebody noticed the name lengths matched, and extracted the letter corresponding to the place in the assembled picture that the missing piece went. Then it was only a matter of repeating this for the other sets and I had a couple people helping me do it, so it finished quite quickly.

At 9:09, others solved 4519, which was the F. And at 9:37 others got 2843, which was the U. Now with all the AFGIMOQUZ filled in, we could read a lot more of the text in the remaining puzzles.

But I didn't immediately see the last two of these, because I had switched over to Shapes in Elder Drifts. I got here just in time to help finish the initial solve of the minipuzzles and read the message extracted from them. It tells us to look at a particular video...

from Alison Burke, whose name I didn't recognize, but I recognized the video when I looked it up. It's a split view; on the right, someone is putting blocks into a child's shape sorter toy, and puts the square block in the square hole first. With the other blocks, after considering putting each block in the matching hole, he instead turns the block sideways and drops it through the square hole, where despite it not being meant to go, it fits anyway, in some cases not fully filling the hole. On the left, we see Alison's reaction, which goes from surprise at the first mis-sorting working to increasing worry and outright disgust as all the shapes are dropped in this one hole.

Somebody very quickly interpreted this the way it was intended. Each answer for the other minis can fit into the blanks in the first mini, in which one space in each row was given such that the givens spell out SQUARE. It matches the given letter and the space or spaces in the answer also match, but there are unused blanks at one or both ends. This let us extract a different message, keeping the words in the same order used for the first extraction.

Meanwhile, people finished one of the other Land of No Name puzzles I had worked on earlier. 3956 consisted of several Fill-a-Pix puzzles. The only text in this one was a single word at the top which clearly was EXAMPLE even before we had any letters, since it introduced a fully worked example of how the puzzle type works. So it seemed like a good starter. I started this and solved several of the Fill-a-Pix, with others joining in to work on others before I was done. We had made the shapes of...

two letters in each grid, sometimes rotated, which led us to wonder whether even the ones that looked like letters in the upright orientation belonged that way. We had 18 total of these letters and a line of 18 digits at the bottom, with a word break and the last four in parentheses, like an enumeration. This suggested they were indices, but how can we use the indices with single letters? It was only when we had more of the letters of the title filled in, MIT Mosaic Lab, that somebody realized these were logos of subgroups at the MIT Media Lab, with the example being the lab's main logo. Those subgroups have names where the letters represent words, providing the missing link.

But that's not all. I didn't really work on 5671, but I looked at it and it was a very simple puzzle. It simply needed enough of the letters to be solved to reduce the possibilities. We reached that point, and it was solved at 10:05, giving the X.

The last puzzle I looked at Friday was The Physics of Linguistic Fracture in Royal Groves. This puzzle features a large number of clues and a small word bank. Even before I looked at it, people found pairs of clues that have overlapping segments. Some of the segments are in the word bank, but many are not. I worked on this until I was too tired to go on. Most of the clues were already answered, but I added missing answers for several and replaced wrong answers that did not overlap in this way with ones that did. I was working on one model for the network of overlaps that was being formed, while another solver used a different model. Neither of us quite had the right model, but our extensive work on the puzzle led someone to solve it at 3:20 AM Saturday.

Also, at some point during this, when there was first something to print, I discovered my printer decided to pick today to experience The False Paper Jam of Death. :-( Fortunately someone else brought a printer that actually worked, so we had printing capabilities.

Saturday

I showed up to open our on-campus HQ at the hour, 7 AM, when we were allowed to begin using it. I took the time to catch up on the overnight activity, which involved:

that the puzzle updates dynamically, only displaying the clues whose numbers end in X in the times whose minutes end in X, and without their numbers and out of order. I've been post-solving this one between segments of writing this report and I'm enjoying it. The clues are fairly easy since they are meant to be mostly solved without crossings, since you don't know where answers go until you have solved a lot of it. Some of them involve what by American cryptic standards are considered lame charades where a compound word is clued using its two constituent parts, but which are accepted in British cryptics.

Atlas of Mosaics consists of a large hex grid. This area is listed as having 19 puzzles in the puzzle list, but those are really mini-metas and actual metas, with several minipuzzles for each mini-meta.

Each minipuzzle corresponds to some outlined region of the hex grid. Groups of minis that go together are located mostly in contiguous areas of the grid. Each minipuzzle has a set of tiles to place in the hexes, without rotation, often with some tiles already placed and immovable. There can also be other markings in the cells, but each minipuzzle is different. And there were drawing tools to let you write notes on the map while solving. Each minipuzzle also has an embedded answer checker (separate from the usual answer checker which is found in the metas). This checker lets you first confirm the grid region is filled properly, and only then to guess the answer corresponding to the minipuzzle. Did I mention that all of this, the hex placements and the notes and the mini solve status, were shared in near-real-time with your teammates?

I also had the role of Chief Cat Herder at 8-12, which was supposed to have the reins on spending our RP to unlock puzzles, manage hints, etc. Someone came to me early in the shift, wanting to request a hint on a puzzle we were stuck on and which had recently had hinting unlocked on it. I told him go ahead and explained how hinting has worked in recent hunts where you provide all your progress, and he did so but got a response in only 2 minutes that we were doing well enough in this Hunt that they were not answering hints for us yet. So that put an end to that.

I unlocked several puzzles for people, and one for myself, Post-mortem Problems in Serpentine Hills, a bunch of bridge hands and commentary on them from which we had to deduce how a team and their opponents did on the hand, but also which commentary belonged to which hand. This went on for a while with me gradually getting more helpers as more people arrived in the morning, until someone got the answer at 10:52.

After a short break to herd cats, I unlocked Electrical Circuit at somebody's request and then stayed to solve that puzzle. There were five images (each accompanied by a number) and a circuit diagram. We all started by identifying the images. The first one...

is associated with a meme in which the quote is often overwritten by alternative versions, but the original is the very peculiar quote, "I do not control the speed at which lobsters die." Pretty soon we figured out that the other images, which didn't have such clear sources, could be captioned by taking a subset of these words, in order. Then the objective was clearly to make the corresponding subsets of the circuit's outputs be true. I transcribed the circuit into logical operators and parentheses, and ...

somebody was coding it up, so I let them do it and broke for lunch. Apparently they made an error the first time through and we didn't solve this until 1:06 PM.

I spent a while after this working on various metas, including Cryptozoologist, which we finished at 1:50.

I spent some time after this looking at the dimensions Hyperbolic Space and Atlas of Mosaics.

I can't say I contributed much to any of the Hyperbolic Space feeder puzzles, but I helped transcribe the data we were getting from the round page, which helped ultimately solve the blank puzzles and the meta in this round. I helped on some of the Atlas of Mosaics puzzles, but there are so many of those and it is so hard to find them in the huge hex grid that I can't really say what it was I worked on at this point.

Later on in the afternoon I got working on Civil Service. By the time I got here...

we had already completed stage 1 of this puzzle, to place the given words in the grid as a diagramless criss-cross, with the capital letters marking all intersections. We'd solved stage 2, identifying the images, determining that each contains ICE as a substring. We'd solved stage 3, fitting the answers in the same criss-cross grid, with ICE in one box, as in a rebus crossword. We'd solved stage 4, overlaying the two solutions, looking through the ice(barns) and extracting from the first criss-cross the message MASYU WHITE AT KGOP SPOTS.

We had trouble interpreting what KGOP meant, but it was suggested this meant just the letters KGOP. The only real explanation we had was that the letters KGOP were to be replaced with white Masyu pearls, and thus create a hybrid Icebarn-Masyu puzzle. The rules are compatible enough:

Form a path with beginning and end, as in Icebarn, since we are given arrows entering and exiting the grid.
The path must go through all arrows in the direction of the arrow, as in Icebarn.
The path must pass through all barns (connected groups of ice), not turning on any ice cell. Since there was only one place in the grid where two ice cells touched, one of those two in a corner, with that one exception the ice cells were simply cells that must be passed through without turning, and crucially, could also be passed through twice, perpendicularly.
The white Masyu pearls follow the ordinary Masyu rule, which is also that they must be passed through without turning, but they must also turn in a square adjacent to each white Masyu pearl.

Two kinds of "must pass through straight" cells with different details made for an interesting puzzle.

The KGOP could be the letters from either grid, but using the first grid failed quickly. Tyler Hinman tried to solve this puzzle, and failed, reaching a contradiction (or "breaking the puzzle" as is commonly said among solvers of some puzzles). Some of us were not so quick to give up on the idea, and multiple of us made our own transcriptions, which revealed that the transcription of the puzzle Tyler had failed to solve had a misplaced Masyu pearl, leading to a different puzzle with no solution. (Tyler says he did not transcribe the broken one, but he doesn't remember who did.)

Just getting the right rules and right transcription didn't make it an easy puzzle, though. We had brainpower roughly the equivalent of a winning World Puzzle Championship team solving it in parallel, some of us in pairs. We broke it more times, but in different ways that told us we were just making mistakes. Eventually, after a break to grab dinner (food helps!), my pair got to a solution. Then after we asked, "What now?", I suggested...

looking at the flavor text and we quickly saw it was to extract from the black pearls, the ones that we weren't given but could have been, which was correct.

I didn't make any progress on the puzzles I looked at for a while until I looked at the Words banner in Fate's Thread Casino immediately after we opened the round at 9:19.

The casino was another dimension, this one with a Gacha mechanic. For those unfamiliar with the term, this refers to a concept in certain video games where you get "loot boxes" with randomized loot. Typically, there are many different items that can come from the boxes and the best ones are intentionally made rarer. Sometimes it is associated with a "collect them all" type of goal, where of course some are rarer on purpose to make this harder. In this case, getting duplicates is worthless. It's relevant for the theme because Pokemon GO's eggs are Gacha-like, with the game selling extra incubators to let you hatch more eggs faster.

The casino gave us chips we could use to buy (or "pull"; think like pulling the arm on a slot machine) clues, Gacha style, under one of five banners. The clues were indeed clues for a puzzle, one of two puzzles in each banner. The clues in the Words banner were all cryptic clues. Two clues per pull. There were two puzzles here, one of which gave us a barred but unnumbered grid and some flavor text advising us about some specific entries by clue number. The other was a diagramless and we were only told to make an 11x11 barred crossword.

We needed to solve some clues first before we could get anything done, and, working with others in our spreadsheet, we helped solve the clues enough to identify three distinct types of clues:

Ones that worked phonetically, using phonetic anagrams based on sounds, and cryptic indicators like "starts of ..." to indicate the first sounds rather than letters.
Ones that involved the phrase "bit that cannot be broken down," which worked if we assumed that was a single letter.
Ones that involved breaking up letters into parts, sometimes with anagrams that worked by parts, and sometimes by breaking off pieces of a letter. This reminded me of 2010's The Capricious Type.

Sunday

I awoke well before campus opened, so after my shower and a snack bar that I had with me at the hotel in lieu of actual breakfast, I went back to looking at Words. By this point, we had solved As the Fly Crows (the one with a grid given) but not made much progress on charfactor. We had solved a bunch of the clues and assembled some of them into the top and bottom of the grid, but what we had assembled looked really bad.

As the Fly Crows was the puzzle that used the phonetic clues. Every clue was meant to work phonetically, and the words were entered into the grid using IPA. This left the "bits that cannot be broken down" clues and clues manipulating letter shapes for charfactor. I worked on reworking the grid for about an hour or so, eventually redoing almost everything we had done on the bottom and filling four and a half rows of the puzzle there. The top I mostly kept but corrected a word and added some words to.

When it was time, I headed into campus, grabbed some actual breakfast, and took care of morning setup at our HQ. At 8, just as my shift as cat herder was starting, we had a group doing an interactive puzzle The Night Swatch. They were making murals to clue phrases which the people who stayed at HQ had to identify by guessing individual words (enumerations of the phrases were given). This involved our Puzzmon, which I had mostly ignored this hunt (you might notice I listed no capstone puzzles in this review; I looked at some, but I didn't make any significant contributions and I didn't do any of the data extraction from the Puzzmon). While I guessed a couple words correctly, my main contribution was in herding the few cats at HQ to look at it.

By this point, the grid was almost completely full, and I was doing mop-up on solving the few remaining unsolved clues, and trying to figure out the extraction. Naturally, I focused on the bits that could not be broken down, and noticed they were always the letters CEORS. I didn't know what to do for that, but it was a key observation, as someone found that the cells in the grid containing those letters formed big letter shapes spelling the answer.

Nick wanted help on Atlas of Mosaics, and I helped him complete a number of its minipuzzles. These were mostly in a section called Underwater Hunt. (Sorry that I don't have links to these, but there aren't links that jump to the individual sections of the map reliably, and just sending you to the whole map is useless. Hopefully, when they put up the static version, this will be better.)

Nick had already identified that it was associated with a puzzle hunt run during MIT's orientation week since 2012 called the Aquarium hunt, whose archives are online (the answers are not, but there is a working answer checker). Each Underwater Hunt mini worked like one of the puzzles in a different year's Aquarium hunt... loosely, since none of them used hex diagrams and some were changed significantly in style to fit into the way the Atlas of Mosaics worked.

We spent hours together working on these minis and ultimately the meta, with breaks to do cat herding stuff and lunch. At one point we stopped working on Underwater Hunt to look at the Atlas of Mosaics metametas. There were 12 of these metas and they contributed to six metametas, two per metameta. At some point I noticed...

that we had two answers in the round that were each two eight-letter words. I thought those went together, and Nick noticed one of them had ONE hidden in it, and the other had EIGHT. This led me to guess they went with the chess-themed metameta called Tree. Other people had already identified this referenced chess openings, and determined the openings corresponding to each line in the tree. I suggested we could use the two answers with two 8-letter words each to label all the pieces in the starting layout. Then extract the letter for each piece that moves on the indicated turns, which was exactly right. Solved at 2:09 PM.

When we went back to the Underwater Hunt meta, we solved it at 3:05. This was also coincidentally around the same time we learned that the coin had been found. But as a team we only redoubled our efforts to try to finish the Hunt.

After this I looked at another Atlas of Mosaics metameta called Playlist. We already knew it had one audio sample of two musical scales and three audio samples each playing the music from The Star Spangled Banner. But I remembered...

that the Star Spangled Banner has more verses. Only the first verse is our national anthem, but there are others, and the three music samples being labeled verse 1, 2, and 3 meant they matched those three verses of the song. Right? I copied the lyrics from a web site. When I listened to the music, I also realized there were extra notes. I started trying to label where those notes were in the song by bolding the words they occurred in.

Before long, my progress was noticed by Todd "pianoman" Becker, who with a better musical ear was identifying which note was played in addition to when it was played. I knew he was much better at this than me, so I let him do that and worked on figuring out the extraction. Other people were working on (and solving) other Atlas metas at the same time, and it quickly became obvious to me that two 8-letter words that each started and ended with the same letter, as in the do ... do of musical scales, belonged here. Ultimately, the extra verses of the Star Spangled Banner proved to be a red herring, and it was only the musical tones of the notes and the order in which they were played, extracting letters from the two answers.

Just before this, at 4:05 we unlocked The Hexagon, the final meta for Atlas of Mosaics. Some solvers immediately identified this as referring to...

France, because they had long ago identified that the overall shape of Atlas of Mosaics resembled France, and a nickname for France is the French version of "the Hexagon" because of its shape. So the list of dates in this puzzle was quickly identified as dates that occurred within the Tour de France in various years. We assembled the legs of the race that occurred on those dates into a virtual Tour de France, connecting them from one leg's end to the next leg's start in the same city, similar to the real race, but this was a tour that had never been raced.

I didn't stay with it, but because of another infamous incident in Hunt history, I sent a note off to Ali, who I knew from the NPL and who I knew was on Cardinality, joking that...

at least France didn't change their districts two weeks before Hunt this time.

That happened in 2016, when a puzzle in a Hunt by Beginners' Luck depended on divisions of France with their own flags, and these divisions changed on January 1, 2016, invalidating that part of the puzzle. Some of the new ones didn't even have flags yet when Hunt started. We wanted to change that section anyway, because three test-solve teams completed the puzzle but none identified the flags associated with France. Once the flags weren't even current, our minds were made up. We searched maps until we found a set of regions that would give the desired extraction, knowing the puzzle was solvable without them and they just needed to be correct.

At 6:39 this meta fell. I had instead switched over to work on puzzles in The Glitch, the final round that, as has been done several times in Hunt history, ties back into the beginning of the Hunt. In this case, we got a glitched version of the Kingdom of the Puzzmon, with parts flickering weirdly. There was just one puzzle for each zone this time, and those puzzles had flicking glitches as well.

We actually unlocked this round at 11:58 AM, but after solving Re: State Diagram at 4:21 we were pretty stuck on other two puzzles we got at the start of this round. I am not actually sure how we were meant to gain access to the other puzzles, but at 5:04, two hours after the coin was found, the rest of the round became available. We didn't do anything at that time to trigger this unlock, so I assume this is when they provided access to all puzzles to all teams, since they expanded the Puzzmon capacity to 100 an hour earlier.

That was the point when I shifted my attention to the round, taking advantage of the Hunt-is-already-over gift. One of the new puzzles was a chemistry puzzle, Chemical Building Blocks, which I worked on with Fuzzy and others.

Because the chemical structures were made of nonsense atomic symbols but which behaved like familiar elements, we quickly figured out this was a periodic table cryptogram. They had given the elements new one or two-letter symbols. The two sets of clues were basically a crossword puzzle on this modified periodic table. The subscripts in the chemical formulas in the reactions were also a cryptogram. I and other people were happily working away at the clues, far from completing them, but we'd gotten the important ones, those that appeared in the reactants, which were all we really needed to solve the puzzle.

Along the way, we figured out the specific alteration to the puzzle that had been made at the glitches. Some of the glitched words were obviously wrong, but context helped us determine what they were supposed to be. Solved at 6:43.

We had already solved Oh My! Interesting Alleles! at 5:32 and it had its own glitched modifications necessary to figure out for solving, as did ¼ Turns, a modification we had figured out by this point, though we needed a hint to figure out the extraction and solve it at 7:10. We also solved Find M(at)e at 6:04 but the people who solved those chess puzzles were confused at our question about how the glitch affected the puzzle. They simply needed to use the tooltips that identified what each glitched piece was supposed to be. This was unsatisfying, but we wrote down the modification to the chess piece images and went on.

Well, we didn't quite get all the puzzles at 5:04. At 6:45, we unlocked snalC ehT tcennoC, the metapuzzle for this round, directly as a result of solving all but two of the feeders, those last two falling at 7:05 and 7:10.

The flavor text told us we had encountered a new Puzzmon, so somebody checked MonArch and found that indeed a new pokemon named Glitchy had shown up there, with all sorts of broken and nonsense data. We tried to match up this data, the names of the categories the data supposedly belonged to, and the glitches and answers to the puzzles in the round, initially only getting a couple of these and some wrong matching. The last two answers helped. Once it started falling into place, it did so quickly. At 7:46, we extracted the answer from this puzzle, which gave us one last task to do...

in MonArch. The whole Glitch was a reference to one of the most famous glitches in the history of video games, MissingNo, an unintended pokemon that can be encountered in the first generation of Pokemon games for the Game Boy as a result of game data getting clobbered and unintended data getting read as the properties of a Pokemon. By imitating the best-known way of encountering MissingNo within the Hunt's MonQuest, players can encounter a MissingNo which gives another puzzle to solve.

We solved that extra puzzle at 8:10, earning an appointment to do the endgame. But not all of us were working on this.

One of the few puzzles unsolved outside of The Glitch after we finished The Hexagon was Da' Bomb, a puzzle about hot sauces that came with a waiver warning us of the intensity level of some of the hot sauces provided. I didn't actually work on the puzzle, but apparently we figured out what we needed from the hot sauces and were stuck on extraction. Around this time somebody figured out how the extraction was supposed to work and had the answer, but we were badly rate-limited due to other people guessing. We used a hint request to reset the rate limiting so we could enter the answer, and did so at 8:03.

We also completed most of the remaining research tasks over the next couple hours. I assume this was mostly remote solvers, since most of the locals went on runaround.

The endgame consisted of two short rounds each represented as a single puzzle on the web site, neither currently solvable with what is there. To The Edge took us to MIT's Borderline Murals. I knew that it was possible to reach the basement of E17 without going outside via a tunnel, but I had never actually done it until the Monster Investigation Team led us there. The murals were added only in 2017 and I didn't know about them at all. When they showed us an image of Spiderman swinging on his web through an orange-and-red portal and asked if anyone knew where it was, I shrugged and replied, "Nope. I know where Wile E. Coyote ran through a wall, but not that."

Almost our entire on-campus team went down there and solved 12 separate puzzles given to us on paper that were based on the contents of the murals. After we finished that at 9:30, we were led back to the Bush room where we were given another 12 puzzles. These puzzles weren't solvable on their own; they required knowing how the previous 12 puzzles worked, and since most people only saw one or two of those, it took us a while until the right people got on the right puzzles.

There was also a main puzzle At the Close. This puzzle referred back to the mechanisms from the entire hunt, so it's all going behind one big spoiler block.

The clues for the answers to this puzzle were all hidden in the style of Land With No Name. There were also hexagons and heptagons for a grid reminiscent of Atlas of Mosaics that were completely hidden.

There were two gumball machines that allowed us to obtain both the letters and the grid pieces in a gacha style, reminiscent of Fate's Thread Casino. We were given four tokens, and each of the twelve solutions to the puzzles we were given here provided 3 more tokens. Together, these allowed us to unlock all the tiles and all the letters of the alphabet, in a random order.

Because there were some heptagons, the grid couldn't be exactly constructed in a plane, as in Hyperbolic Space, and it was meant to tile, or probably close up on itself in a sort of sphere, also as the solution to that puzzle did. There were connections on the "edge" to other parts of the edge, marked with sets of 3 colored blocks, as were also used to help match the interior.

The answers to the clues did not fit in the spaces for some answers, because they had to be modified again with the glitch mechanisms. Once we did, they could fit on paths we could find through the tiles, with the bars between letters of the answers matching ones on the tiles. Unfortunately, we only had one copy of the paper tiles, and there was no way the 70 of us could all participate, so about 15 people were working on it, and everybody else chatted, did stuff online, told their non-participating friends and family we were almost done with the Hunt, etc. Some people even played the Puzzmon card game we got from one of the other rounds, one of them describing the actual game as "broken as heck."

Each clue answer had an arrow at the end of it, showing the approximate direction to turn at the end of the word to reach the next word. Because of the disconnected map, we didn't do this right for all the connections the first time; also, people naively assumed we would just read the letters pointed to off in the order of the clues, which were alphabetical by unglitched answer. Once we extracted all the correct letters, we realized they were pointing to the starts of other words, and gave us a sequence for the words, and then people used the otherwise unused circles on some tiles to extract an answer reading along the sequence of answers.

We could have done this a lot better if we had had somebody transcribe into our spreadsheet what letters were on and adjacent to each tile, along with where the groups of colored blocks were. More people would then have been able to participate, including remote solvers, and we would have caught errors more quickly. But nobody out of the few who could even see the things was organized enough to do that. If we were racing against another team to win the Hunt, we would have easily lost the race because of this.

Monday

Monday morning, teammates finished the rest of the research tasks by 9:11 AM. We also solved one remaining level (minipuzzle) within the Atlas of Mosaics to achieve a 100% complete of everything, but only at 1:08 PM just as the wrapup finished.

I went to cleanup at our team HQ, and then wrapup, and chatted a bit with teammates, including handing out a few leftover 2023 coins to people who still didn't have theirs.