MITMH24: A Skymin's Perspective

Last updated . Moved page to a new filename. Also a little formatting to align with the rest of the website.

As of this writing, the public access version of the hunt website is available, so the puzzle links are working. They don't yet have a permanent home, though.

Previous updates

. I posted this and it made quite a discussion. I expected the discussion and I think it's worth generating, but some people mistakenly assume what I'm saying is also official UE policy. I'm trying to plaster a lot of disclaimers to say all these is just my opinion.

. New page. I found I wanted to style things more than Markdown offered, so I decided to just write it here instead of on Cohost.

MIT Mystery Hunt 2024 has just concluded, thus ending another puzzling year. (Or perhaps starting it?) For the third year in a row, I hunted with ⛎ UNICODE EQUIVALENCE, a 37-member team of tight-knit friends that are also strong hunters. We finished third through the long hunt, less than an hour after the top place.

1st: Death & Mayhem (5:21 AM)

2nd: The Providence Planetary Stations (5:31 AM)


4th: ✈✈✈ Galactic Trendsetters ✈✈✈ (6:53 AM)

5th: Cardinality (7:44 AM)

Congratulations to Death & Mayhem! I can't wait to see what they will cook up after a gorgeous hunt in 2018.

The rest of this post is about my thoughts of the hunt, including a recap of what I did in UE, my favorite puzzles, and my parting thoughts.

If you want to leave a comment, do so on my Cohost post about this page. (There's also discussion in a thread in Puzzlers Club, although it's discouraged as it usually spills to other channels in the server.)

The public access version of the hunt website is available at However, you have to first log in by clicking the "public access" link, otherwise all puzzle pages will return a 404 error.

In this recap, puzzle titles are written in bold and in brown.

Warning: A few links lead to It is an imageboard of furry pictures; specifically, a safe-for-work version of (which has adult content). Despite being safe-for-work, you will still see an "over 18?" prompt on entering e926. All the links provided here are safe; they just lead to cute Pokémon pictures (in the recap below, I explain why there are these links in the first place). As such, it should be safe to click yes on the prompt. Despite that, if you feel uncomfortable anyway, the links are not important by any means, and you can skip them.


There will be spoilers for the hunt as a whole, including puzzles and metas that I looked at. None of them will be marked.

I'm leaving a huge empty space so that you can turn back before you scroll further.


DISCLAIMER: This is my opinion, my piece. It is a recap of the hunt, but it is as seen by me. My opinions may be interspersed with the recap, and those opinions are mine. They don't reflect the opinion of any other people or any team, not even UE's.

Note: Times are in SGT (Singapore Time), as it's my time zone. SGT is 13 hours after EST, so the Friday noon EST kickoff is actually Saturday 1am for me. Days of the week refer to the 12–15 January (Fri–Mon) weekend. You can hover over a time to see its equivalent in EST.


There are two goals for UE. First, finish the hunt; second, don't win.

I suppose the second one wants a little explanation. As a team, we tried writing some sort of UE hunt a couple of times in the past. In all cases, we got burnt out very early on, with the hunt nowhere close to being ready. We have a motivation problem. Since the winning team of MH is tasked to write for the next hunt, we decided we're not ready at all. Maybe in the future, but not now.

As part of the second goal, we decided we would handicap ourselves in some ways. From what I see, UE appears to have these handicaps; some natural handicaps, some artificial if we ever got in danger of winning:

UE as a team is also fairly tight-knit. It is supported with a lot of remote solvers, and I see these features in the team (and its server):

In short, we're well-equipped to make sure we have fun while still being a very strong team.

Hour 1

We all gathered in the #lounge voice channel to watch the kick-off, although technical issues meant the kick-off was delayed by around 15–20 minutes. When they said we were banished to the Underworld, we were excited at the prospect of a 2007 theme repeat. The safety video was hilarious, although not to take away from the other actors which were also funny. Okay, so our theme is to escape the Underworld. Time to wait for the puzzle release.

We weren't sure whether the puzzles would release at 2am as promised, but surely the delay wouldn't be more than 15 minutes tops, right?

The first problem immediately surfaced. At first I could load the website just fine, but when I entered my team's login details, the site just stopped responding (it got stuck in "logging in", with all buttons disabled). Then I tried to refresh the site, and it just fell apart entirely, giving 500 errors. Uh oh. Hopefully it's just the site being under heavy load due to everyone wanting to log in, but also, everyone is going to want to keep refreshing the site.

The promised time for puzzle release came, and the site wasn't up by any means. And then 2.15am. Then 2.30am. We were itchy for puzzles, although we also recognized the tech team had to be working really hard behind the scenes and that we couldn't do anything. So we just waited. The Singapore contingent (Level 51 and lovemathboy) decided to go to bed, although I decided I would stick around for a few hours because I expected the initial rounds to fly by quickly and I didn't want to miss story beats.

Around this point, we realized there was puzzle content behind our nametags (okay, at least the on-site people), and we started compiling as much information as we could do while the site was down. We did all we could, and yet the site showed no signs of getting back on. We started streaming some games: three people were playing vivid/stasis, and we could see elara on-site playing Phigros. Hey, there was nothing to do.

At 3.15am, we received an e-mail saying that they would release an offline copy of a round while they were fixing the issues. Well, they said they would in the next 15 minutes. The offline copy didn't came at that time; instead, at 3.50am people said they were able to access the hunt website and see puzzles, so we were off to the races. Only later at 4.10am that we got an e-mail saying that the site was up intermittently, and while they were working on site stability, they also released an offline round. This does mean we had a massive width, so we could jump to whatever we liked. (We also received a HQ interaction that explained, among other things, that they "fixed" the site by making it crash faster so it would reboot faster. I thought it was a joke, but now I'm not so sure.)

Ultimately, a very unfortunate start to the hunt, but it was better than nothing. I recall the 2016 hunt had even worse tech issues that every round became fully offline; at least this time the site ended up mostly online barring this initial hiccup. We had puzzles, that's what's important.


Technically this was Saturday for me, but it's before I slept. I ended up sleeping at 6.30am, so I had approximately 2.5 hours of puzzling.

The first puzzle I looked at was Why The Romans Never Invented Logic Puzzles, which was a Kakuro with a twist. A few people (I believe ManyPinkHats and TheGreatEscaper, at least) were already working on this, with ManyPinkHats streaming his solve. That meant he couldn't see the text channel, though, so I had trouble forwarding my observations in the chat.

The puzzle was tough, and while we started off pretty well with one quadrant solved, we were pretty slow in getting a second quadrant. I also realized the long entries would be really awful to think about, so I started coding a program that, given the clue number and the number of cells in the word, would spit out all possibilities for it. It was helpful, even though "840 in 9 cells" gave 769 solutions. At least we could filter them out based on the known cells.

It took us collectively 1.5 hours to solve the puzzle, even with the tool introduced around halfway through. It was hard. (Pretty fun idea and puzzle, though.) We got the final clue phrase, and I made the aha that the cluephrase meant we should convert the letters using A1Z26 and then to Roman numerals, and then we had to divide the sequence of Roman digits in a new way to get an English word. (This is also confirmed by the two strings extracting to the exact same sequence of Roman digits — which, now I think of it, helps error-checking — and that it's elegantly reusing the same concept from the puzzle.) I helped with extraction, wow! We got the answer, and there we go.

During this time, I also briefly looked at other puzzles. I saw Roguelikes with a K, which seemed interesting (game-based puzzles are often fun!). But I was busy with Romans, and other people have picked it up in the meantime, so I decided to leave it and go look for another puzzle.

I saw Do You Like Wordle? unlocked. I'm not that good in Wordle, so I just announced it before moving elsewhere. In theory, it seems like it has a black box component.

I then saw a chess puzzle in A King's Best Friend, so I quickly jumped in. At this point, moo was looking at the puzzle, and gave the observation a cerberus would likely be able to move in three directions, but didn't have an aha yet. I pointed out it was likely the three directions given by the triangle (diagonally forward or immediately backward), with some limited range that we concluded to be 3 cells. (The directions were also clued by the puzzle title featuring 🐶s.)

In addition to all that, I also noticed the puzzle hinted at FEN in several places, and that's when I realized every line formed a valid FEN row (because it would fill exactly 8 cells). Solving every board and taking together all the answers gave us one final board, and we solved the board to get a move string.

The ending was a bit more difficult for me; in a way, it was more of "usual puzzle hunt" nature of deciphering what on earth this could mean. It took us around 10 minutes to crack this, and I'm glad I didn't have to call people for help on extraction.

It was 6.30am. I noticed we had a new round open on the website, and it was the same round as the offline round we had, so I pointed it out before going to bed.

One last thing: we had a #salutations channel where people could sign in/out (e.g. when they wake up or go to bed). I decided to leave a cute Pokémon picture with each of my messages, so here's the first.

Saturday early morning

Good morning! Well, it was 2.30pm. At least I had enough sleep. I ordered food delivery so I wouldn't have to be distracted from puzzling.

Reading after the fact, it seems I missed the mid-hunt runaround that took place at 8am (about an hour after I actually fell asleep). Whoops. I woke up to a bunch of new rounds that somehow took place in the US. I guess that's all for the Underworld fun, then. (Also, rounds across the US felt a bit reminiscent of the 2022 hunt, although repeating an idea is not by itself a problem.)

Retro Chess Puzzle was solved when I was entirely asleep; I would have loved to do the puzzle, although I can just do it after the fact. I also missed the entirety of Hell round, which was a shame. I likewise missed Hall of Innovation round in 2023, and I would have loved to get the "oh my god" reaction together with the team. Oh well, maybe next time. D&M, hope you're putting a bunch of gimmick rounds!

The first thing I jumped into was Mandalay Bay, although I didn't contribute too much to it. I did make sure the sheet was kept tidy; three of the emojis were very new and not entirely supported on all systems yet, so we used conditional formatting on those, and I had to make sure people were pasting only plain text instead of copying formatting as well. I'm not claiming much at this; lovemathboy, boboquack, and TheGreatEscaper were the biggest contributors.

Solving that unlocked Sahara, which was very much a logic-like, and we quickly jumped on it. Someone else made the aha that we could figure out what numbers each person bet on. After that, it was largely a logic puzzle, and all of us that gathered for the puzzle contributed in solving the puzzle, spotting deductions and double-checking everyone's deductions are correct. I can say I got an aha, after figuring out how the final extraction worked (it was a roulette setting, so I tried plotting the winning numbers on the betting table; then I recognized clues in the flavor text to confirm this). It was a collective work and I enjoyed the puzzle.

Next up was Flamingo. Another logic puzzle! Quite a lot of puzzles in my wheelhouse. Of course, it's not a hunt puzzle without some trick; we were wondering what on earth "E" meant, and having a 3-cage to multiply to 181 didn't make sense. Someone made the insight that it was hexadecimal digits, and I pointed out each ring (in addition to each face) had 16 cells, and we were off to the races. As before, we just made sure each step of logic was valid.

The collective work continued, although it's no longer a logic puzzle. Schrödinger's Maze got some work already done, although mainly just to figure out how the mechanics worked. Solving for each hidden word and each hidden grid still took quite some time, and I remember doing one of the hardest grids (grid 4, with 13 blocks) and finishing exactly as the others finished extracting the final answer. Whoops. It's still a fun idea and puzzle.

All these were done in a few hours; it was 7pm when we finished the Maze. I think, at this point, we unlocked The Champion. I recognized the Magic: The Gathering reference, but couldn't figure out what else it was, and I stepped away for a little break. Solving several puzzles in succession was gratifying, but I don't want to be too burned out.

Saturday late morning

I came back at 9pm with no new insight for The Champion, so I largely bounced between puzzles with no avail for a couple of hours. I know my strengths and my weaknesses, and sometimes the open puzzles just don't give me much to work with, so it's a good time to step back and take a little break.

At 11pm, Isle of Misfit Puzzles was felled, and with that two new puzzles opened: Scan and Sumantle. I instantly latched onto Scan, and... everyone else went to Sumantle.

chao: everyone wants the other open puzzle -- lmb: this puzzle ain't one for collaboration, have fun coding

Well, all for me, then. I coded a factoring program specialized for this in around 10 minutes and extracted the pair of QR codes. I interpreted the message, took the XOR of the other QR codes in the pairs, and got a message saying "Braille at bottom". This took me a while to figure out, though, because I was looking at the bottom of the QR codes that I XOR-ed together. I sent a call for extraction help, but barely a few minutes later, I realized the first QR codes were very unconstrained (one-letter message in a pretty large QR code?), so I found the Braille letters I was looking for and I got the solution. I entirely soloed a puzzle, and it's not a logic puzzle either! Since I soloed the puzzle, I took a bit of time writing up my approach and solution in the puzzle's channel, so anyone else could look at it. (I also dumped my code there.)

And then Infinity Book opened. I poked at it a bit, but ffao made a realization in The Champion that it referred to The Theriad, so I switched there. We put in some time matching cards to lines... and then got stuck again. I don't know why we never actually made a connection to The Iliad as well, but this shows my weakness a lot: I have no hope with research-based puzzles.

I turned back to Infinity Book, where lovemathboy and SeptaCube cracked the initial encoding portion, getting a mapping from each character to its two-letter encoding. So it's time to decipher the book. The first page gave us a message... followed by a garbled text. Deciphering that gave us another message, and another garbled text.

Oh, it's Infinite Cryptogram time. The first thing I checked was that it handled high page numbers accurately, because part of the problem in Infinite Cryptogram was that it only accepted integers that could be represented in double-precision float. (Well, the other part of the problem was that the encoding was to words of differing lengths.) Luckily it did, so crisis averted; this puzzle, in theory, should be easier (and more reasonable) than Infinite Cryptogram. The initial message was always 50 characters long (a shorter message was padded with space), so we know the first message is given by the first 100 characters (encoded once), the next one by the next 200 characters (encoded twice), then 400 characters (encoded thrice), and so on.

I set up a quick program to decode stuff. The instance of Infinity Book I first saw had [9] on the bottom, so if that meant the 9th layer, I figured, I would only need pages up to 52 — a bit far, but I could easily take the pages manually, and it would serve as a good check. Unknown to me, there were several different instances of Infinity Book, and SeptaCube looked at one with [117] on the bottom.

After getting the realization of the round structure, we decided to handle the 9 first. We got the answer word and submitted it to confirm we were on the right track, and we brainstormed for ideas for the 117.

Something easy to do is to take a character, and always take the first letter in the encoding. Alternatively, always take the second letter. Someone in chat said, neither of these by themselves was unique. I might have misremembered it as, these taken together to form a pair wasn't unique either. So, with nothing else in mind, I decided to simply do a random try. My program would generate a random path of taking first/second letter each layer, and would come down to a certain page, for which it printed the page number. I would then take the page there and fed it to the program, which would process it and eliminated letters that could no longer be candidates. I tested this with our layer 9 solution and it seemed to work, so I went for layer 117.

It gave us the answer RETAILER. We submitted it. It was wrong.

SeptaCube, while at it, had also set up a program as well (I think it's some sort of computation on the Google Sheets we were working on), and came up with the same answer. So we weren't sure where we went wrong; we suspected an off-by-one error somewhere. That's when Max came to our channel with an @everyone ping:

Max: stop working on this puzzle. the puzzle has a problem and we will be informed when it is fixed

Well, I suppose the good news is that it's not a bug with our program. It's also interesting that we got this message so soon after we submitted it; there's a chance we're the first team to see this puzzle at such a deep layer, which suggests we're going strong.

Either way, we talked a bit about this, and Max gave the go-ahead around 10 minutes later. We got a completely different word, which is a bit of a shame as we have to redo the work, but at least this was marked correct.

We also now knew about how the round worked, and we bashed at every instance of Infinity Book we could see. SeptaCube moved on to other puzzles, and I also looked at other puzzles, but I kept my program handy so that I could solve further instances as they appeared. (Part of it was because I was coding directly in the Python terminal, and my code was so ugly that I wasn't sure I would be able to write it again quickly. I also tried my best to stay up until the round's meta was solved.)

I also felt this random approach (remember, my program queried random pages) felt weird, so I tried to analyze the structure of the encoding. I double-checked the idea of always taking the first/second letter earlier, and realized my assumption was wrong. At least for the lowercase letters, the pair was unique. Then every character could be encoded once, and then each of the lowercase letters could be expanded in constant space. That made it less random — a word with length N could be solved by 2N+3 queries — and I decided it was good enough. The code was still very dirty, but there wasn't much time to be gained by improving my code when I could work on other puzzles.

It took me around 2.5 hours working on this puzzle. The idea was fun, but I'm not sure I liked how I approached it. Well, whatever, a puzzle is a puzzle.

Saturday afternoon

There was a gap of an hour — I may have showered at this time — but I went back puzzling around Sun 5am. The puzzle was Realize It, and I quickly read the rules and contributed on the solve (again streamed by ManyPinkHats). It was a weird genre, and it took us a good while to wrap our head around it. We eventually extracted the clue phrase, though: ALLOWABLE SEQUENCE. "Keep going", the website said, when we submitted it as an answer. Huh?

Turns out allowable sequences were a thing. I found a paper, which had a definition of allowable sequences, and that was enough. We mapped the desired configuration of points, and... then what? Extraction, the bane of my existence (and many others). I left the puzzle in the hands of Deusovi, which solved the extraction in 10 minutes.

I think the puzzle was fascinating. It introduced me to a new concept in math, and made good use of it as a logic puzzle. My kind of hunt puzzle for sure.

I tended to more Infinity Book submissions until Hydra was solved. I then dumped my code to the channel for preservation before closing my terminal to save memory, and so I could start fresh to deal with Marathon Block Pushing Game.

A few people started poking at it, but I immediately knew what I had to do. I explained the concept of the gadget to Deusovi, including the fact that the whole thing was basically a huge binary counter, and he decided I was better equipped at doing it as he left for another puzzle. I did some coding to compute the number of steps required, matching it to the debug information provided in the PuzzleScript link, and got the solution, telling me to BUILD SOKOBAN.

Well, the second part of the problem came up. I remembered there was a quadratic instance for 1-box Sokoban, but I couldn't find the construction online, so I had to figure out how it worked myself. Fortunately, I came up with the idea of pushing a box backward through one-way gadgets that otherwise forced you to move forward, and I whipped up a level quickly. Pretty ugly, but it did the job with the solution taking 1763 steps (I needed 1000).

I love hunt puzzles that are secretly math/coding problems like this. (Logic puzzles are basically similar.) The rules of the game are usually pretty clearly laid out, and it's about being smart, instead of slogging through a lot of research. Some people might like the latter kind of course, just that I don't. And being in a smaller team like UE means I get exposed to many more puzzles – the width of the hunt remaining massive at around 25+ open puzzles also helps – so I can often find all the puzzles I want.

It was 8.30am; I felt tired after powering through the Sokoban coding, and I signed off.

...or did I? I looked at A More 6 ∪ 28 ∪ 496 ∪ … very briefly. I noticed the constants 118 and 92, and claimed this was a chemistry puzzle, before heading off to bed. (Jonah then recognized what the puzzle actually was: the U.S. Congress. Someone should make a puzzle using the fact that there are 118 synthesized elements and 118 Congresses so far.)

Sunday early morning

I woke up at 2.30pm, around the time as the on-site contingent was getting some well-deserved sleep. Honestly, I don't think I had enough sleep, but I decided to wake up anyway. Unlike yesterday, I decided to head out to get some food; the puzzles could wait for an hour or two. (I haven't even properly signed in with a Pokémon picture.)

Well, turns out Queen Marchesa to g4 unlocked. I looked at it briefly and figured out this would be a puzzle I'd love to do. Hopefully I could still do it after I returned back...

...or did I have to wait until I got on computer again? I solved one grid in my head and decided I could actually contribute a lot here. moo was on the call and was managing the sheet, and ffao was also present with some help. (moo knew chess well but wasn't familiar with Magic at all, so I explained the basics. I was fortunately in the intersection of Magic fan and chess puzzle fan.) We knocked off grid after grid, even as I enjoyed my hearty lunch of nasi goreng and lotek. We solved the puzzle before I got back on my computer. I love puzzles that I can solve on the go.

So I got back, properly signed in (caution: anthro), and looked at the open puzzles. Jigsaw Sudoku was just solved by lovemathboy and SeptaCube (although I noticed they didn't put in the answer to the channel, so I did so), and it opened a new round including Jigsaw Slitherlink. I thought of hopping to it, but lovemathboy was handling it better than I could, especially once I found out we would have to resize the pieces (I don't think I have good image editors for that resolution), so I left it for Fairly Thorough.

There was some obvious work to do here, and I spent a good hour or so making a grid of all the rooms. I couldn't make sense of what it was, though, and I left it for other people to chew on. (To my knowledge, very few others ever gave it a look.)

At this point I also felt incredibly tired, and I decided to take a nap. Perhaps it was because I didn't have enough sleep earlier, but my nap ended up taking three hours.

Sunday late morning

I saw it was 11pm and cursed why I ended up sleeping so much. At least I still managed to solve a puzzle earlier, and I started looking for our current puzzles. It seemed there was little progress at this time.

I don't think I contributed too much this session. I mentioned that nobody was looking at Make a Winning Hand, so I played around a bit and also called lovemathboy to take a look (he likes interactive / black box puzzles).

I then noticed there was Puzzle with a Twist, and at least I could tell what I had to do. Obviously I had to assemble the six faces into a Rubik's Cube. I started with the first cube and did it manually, before deciding I could code a solver for this assembly. (It just tried every possible rotation of each face and checked that every valid piece appeared exactly once.)

I got all assembled cubes without much trouble, so I started actually solving them. The first example solved without any hitch. The second example also solved; I saw there was a rotated center, and I assumed the 1 meant 1 facelet was wrong. The third example couldn't be solved by an online cube solver, but I managed to find a three-move solution that gave everything correct except for one flipped edge. So the 2 meant 2 facelets were wrong.

And then it went to disaster. I looked at the fourth example and I couldn't figure out any way to solve it with just three moves — the previous three examples all only took three moves, so I figured it surely must hold, right? I did notice it was solvable by rotating a corner piece, but that didn't explain the 4. There would be only 3 facelets wrong! I couldn't figure out what went wrong, and decided to drop the puzzle as I looked elsewhere.

On retrospect, I actually looked up the mathematics of Rubik's Cube earlier. I found out there were 12 orbits of the Rubik's Cube group (2 ways to flip a single edge piece × 2 ways to swap a pair of edge pieces × 3 ways to twist a single corner piece), and if orientation of the face centers was taken into account, there was an additional factor of 2 (a center might be rotated 90 degrees). But it didn't register to me that the product of everything was 24. It didn't come to my mind that it was counting in some base. With how the 0, 1, 2 matched the number of incorrect facelets, and the fact that the solutions were all three moves, I thought it had to be that way and I never looked for an alternative explanation.

I'm not sure how I feel about the puzzle. Maybe I got skill issued hard. Maybe it could be presented better. It was a fun idea, at least, and doing some coding to solve the assemblies was interesting. I just wished I could have brought the puzzle to completion.

I dropped the puzzle in favor of IO. I normally balk at research-based puzzles, but maybe I can be more helpful when the research is about technology? For example, I noticed that the constant 0xEBCD1C0000 had "EBCDIC".

...well, we still got stuck. As I'm writing this, I'm checking the puzzle page. While they don't have the solution yet, they do already have some canned hints, and I don't think we even got the first step mentioned by the canned hints. I guess I really suck at puzzles where I don't know what I have to do.

Sunday afternoon

At Mon 4am, we had a surprise visit. HQ was coming with an interaction, and they brought... free answers? We got three of them, in fact. And we had to use them right away. The team gathered immediately: do we want to use them? Where are we using them?

There was some opposition, but we decided: given our current state (we had 6 rounds open, and we hadn't seen the meta of 4 of them), we were actually at risk of not finishing. In addition, we got the vibe, from the people that came for the interaction, that we weren't at risk of winning. We wanted to optimize fun, and that included unlocking more puzzles (possibly giving something new to chew on) and metas (we might not want to win, but we wanted to finish).

So we proposed three puzzles for which nobody wanted to work on. The first puzzle we proposed was actually written by someone that gave the interaction, whoops! ("I can tell you the answer to that without having to look it up. And hopefully you'll give it a try later to appreciate the author's ingenuity and handsomeness." Okay, not exactly those words, but something like that.) That puzzle was also titled Time for a Drink!, and we joked how that meant we skipped the hydration check, just like some people backsolved Remember to Hydrate! in Teammate Hunt 2021.

Once the interaction concluded, I immediately went to look at our open puzzles and prepared a few options for puzzles to be berged next time. Others were welcome to propose some puzzles too, but we ended up mainly using my list because others were busy puzzling.

We unlocked The Farm and I made a Desmos graph for the big equation. Not sure what else to do, though, so I left it as I jumped to another puzzle. I saw Look Before You Leap was stuck on extraction, and I got a big brained insight that some of the questions could refer to whole objects rather than just single letters: it was a whole LOG behind us on the 26th step, not just the letter G. That gave us the answer, and I moved on.

Island Hopping unlocked, and I immediately worked on the Nurikabe. I also noticed words outside the grid, though, and I figured those might be clues that needed solving; I streamed my solve of the Nurikabe while asking around for anyone to do the clues. The Nurikabe was pretty tough; it took around half an hour, with a couple of intuited guesses, to fell it.

Around 6am, in the middle of my Nurikabe solve, there was another HQ interaction. I was dealing with the Nurikabe, and my internet also dropped for a few moments, but I got news that they used my list for a puzzle to be berged.

With the Nurikabe solved, Deusovi joined as we tried to figure out what to do with these words. Turns out they weren't clues; they were just outright entries to be slotted in. Deu made the insight of what the numbers meant, and we fit all the words together. We were having trouble with extraction, though; my suggestion about "criss-crossing" and "islands" to mean to look at word intersections on islands didn't give anything, and I left it for another puzzle. (And then Deu solved the extraction soon after.)

At this point, I recognized we would need more puzzles to be ready on the chopping block to be berged, so I prepared a new list. Nobody else touched Puzzle with a Twist, so I put it up on the chopping block as I also didn't know what to do with it.

12pm came; the hunt was going for almost 60 hours (okay, with the technical issues, it's closer to 57 hours) and we were nowhere close. The team wanted to start using hints on puzzles that were particularly close, such as ENNEAGRAM where we were missing only the extraction. As before, people were initially a bit hesitant, but we decided it was for the better. (I believe, the moment the hint came back, everyone groaned because it should have been so obvious. In my defense, I didn't work on the puzzle and I very much didn't know the Enneagram existed.)

I signed off, but I was still checking on my phone. We had just solved Duet — with an answer that was unfortunately not BOHEMIAN RHAPSODEEZ NUTS — and we unlocked A Night at the Opera. I saw our working sheet had some chess pieces mentioned, so I pointed out there was a famous chess game known as the Opera Game, with a note that whether it was "related or not, idk". (It was very related, and we got the answer half an hour later. We also learned this was the first solve of this meta, about 2 hours faster than the next team. All only because I happened to know this game existed, so we didn't have to research based on clues scattered on the page.) I can claim I helped solving a meta in a big way, now. And with that, I actually went off to bed.


I woke up at 6.30pm to the news that the coin "will have been found". This is the same wording as last year's, when the last puzzle was solved but the campus was still closed. Along with that news, hints for metas were opened. We had one final meta left, Aphrodite's Plutonic Affection Connection, and we weren't making any sense with it. We identified the anime characters, the Persona-related answers, but we were missing their connection.

I decided to go out to get some takeout dinner. The team was itchy and wanted to hint it. As I was getting ready, though, and as the team was writing a hint request, Ricky ended up getting an inspiration about love triangles — and hence finding the third side of a triangle — and we ended up solving the meta.

The final puzzle, Road Trip Redux, opened up while I was queueing to order food. It fell in around 20 minutes; I hadn't even returned back to my computer! At least I caught the moment where we finished. At 7:07:27pm, we finished the hunt after approximately 63 hours.

Of course, with the hunt finished, we had nothing else but to pick on the remaining puzzles, although I didn't find any I wanted to do. I mostly just waited there until the runaround at 11pm, still feeling high after the hunt. The runaround came, with someone streaming it for us remote hunters to watch. Then the wrap-up came, and another Mystery Hunt concluded.

And of course, fun tidbits came after that. In Puzzlers Club, we learned some things. Perhaps most notably, I summarized it here:

UE was 4 times smaller than D&M (37 members vs 130+). UE used 10 times fewer hints than D&M (we used 4 hints; other teams used hints quite liberally). And UE still finished only 40 minutes behind D&M.

UE also finished ahead of ✈✈✈ Galactic Trendsetters ✈✈✈ (45 minutes after us) and Teammate (didn't finish, missing two metas). Thinking about it, it's kind of scary what a powerhouse my team is.


My favorite puzzles of the hunt! Well, sort of. The first section is for puzzles that I actually touched during the hunt and so I could tell I enjoyed them. The second is for puzzles that looked interesting, but I missed them one way or another, so I couldn't say whether I'd enjoy them or not. Obviously, as this is my list, it's very much geared to my interests; I'm sure other people will also post their favorite puzzles, so you have plenty of resources. Still unmarked spoilers here, though, although less spoilery than the recap above.

Each list is sorted by appearance on the List of Puzzles page.

I did these

Why The Romans Never Invented Logic Puzzles: Logic puzzles with a twist are a staple in hunts, and they are often interesting to solve. The twist will often bring something unexpected that makes you think about the puzzle in a way you never did in usual logic puzzles.

A King's Best Friend: Chess problems! I'm usually not the best at usual chess problems, but generally hunt puzzles tune down the difficulty of the problems, so that they are more gettable — not everyone is an expert chess problem solver, after all. That makes these problems enjoyable, once you figure out the gimmick. Also, it has a final step that I've never thought about before.

Flamingo: More logic puzzles with a twist! For this one, I think the puzzle itself is pretty tame (although still fun as a logic puzzle); it's figuring out what's going on that's the novel part.

Mandalay Bay: Black box puzzles are fun. What's going on? It's always a delight to figure that out, and usually black box puzzles involve minimal research, if at all — everything's right there. On retrospect, I think I end up not being too attached at the puzzle, but it was still pretty enjoyable to tinker with it.

Sahara: Although full of text, the puzzle is actually a logic puzzle (so I guess it's more of an Einstein-like). There's some research needed, but if you get over that hump (or someone helps you with it), the logic part was pretty nice. I won't recommend this as highly as other logic puzzles here, but it's still something.

Infinity Book: This puzzle is special in a round with a special structure, since this puzzle exists in many different instances. All of them are the same except for the number at the bottom, in which the first five have very low numbers (3–11) and all the rest have very high numbers (101+). It's ultimately a coding puzzle, or more generally a puzzle where you're given everything and you have to figure out a strategy to solve it. In a way, it's a more broad version of logic puzzles, and this kind of puzzles also appeals me. The puzzle itself has two parts to it, which I'm not sure I'm particularly fond of — they feel detached from each other — but I still think it's interesting to work through it.

Realize It: Logic puzzle! Except it's in a genre completely unlike any other I've seen, and it somehow relates itself to a new math concept I didn't know before. The logic was tough but fun, as if discovering new deductions in a new genre, and the final step is neat. I struggled with extraction here, but I think it's pretty fair, all things considered.

Scan: I think, for a puzzle at this point of the hunt, this is surprisingly small (I soloed it in around 20 minutes). But the important thing is that it's fun. It's a coding puzzle, because you're not going to factor a 881-digit number without some clever stuff going on.

Schrödinger's Maze: Black box puzzle! It's always fun to experiment yourself and figure out what's going on. Perhaps one problematic thing is that it expects a real word as input, which gets in the way with your experiments, but restrictions breed creativity, and figuring out how to handle the restriction is pretty interesting, at least for a while.

Marathon Block Pushing Game: It's a coding puzzle. You basically know exactly what you have to do; the challenge is to actually do it. And... surprise, when you solve it, it gives you a new task to do. It's two puzzles in one! I have personal attachment to this, because I do a lot of mathematical research on games, and I once gave a talk to Puzzlers Club about complex things and gadgets in Sokoban. So everyone gave me this puzzle to do, and it was delightful.

Puzzle with a Twist: The puzzle idea is interesting; it feels like not a usual kind of Rubik's Cube-based puzzle. Although, as I complained in the recap above, I feel there are some red herrings as you're trying to figure out what's happening. I still appreciate the subject matter regardless.

Queen Marchesa to g4: This is probably my favorite puzzle in the hunt. Chess puzzle? Magic: The Gathering puzzle? Some hidden mechanics that you need to figure out, much like instructionless logic puzzles? Tricky problems that are wholly self-contained and fun to solve? All check.

Island Hopping: Despite the words, this turns out to be largely a normal logic puzzle. Well, two logic puzzles to be done sequentially, you can say. The words don't need to be "solved" or anything, although figuring out what the heck are the numbers for is a bit tricky (and so is the extraction). I guess for me this is so-so, but logic puzzles are always appreciated.

I want to do these

Do You Like Wordle?: I'm not a big fan of Wordle, but I am a big fan of black box puzzles, and this sure looks like one — you're not going to solve 30 simultaneous Wordles in 6 guesses, after all, there's going to be a trick.

Roguelikes with a K: Puzzles that give you games to play are always fun. (Well, they are a special case of interactive puzzles.) I only looked at one but it seemed puzzly, and I think the others are fascinating on their own as well.

Steam Library: To be honest, I haven't really done much meta-matching puzzles in general. But they seem to be fun. They are always a mixed bag, because once one of the metas needs some research, it's going to be painful for me. But that's why you solve these in groups. Also, theming the metas after games adds some fun factor.

Retro Chess Puzzle: Chess problems! Retro chess problems, while at it! I'm a big fan of them and I'm so sad I was sleeping when this puzzle got unlocked and solved. It looks incredible. I spoiled myself at a trick that happened later in the solve, and it looks even more impressive after that.

Hell, MI round: The entire round is gimmicked and so I consider them as a single cohesive puzzle as a whole. Gimmick rounds are always fascinating to see; one day I'll get the chance to experience them in real time.

99% of Mystery Hunt Teams Cannot Solve This!: Clearly related to math, and it seems like it dives deep into all the mathematics. I think ultimately it will have a lot of research too, but liking the matter at hand does make it more bearable to do the research, so I'm interested in giving it a try some time.

The Hermit Crab: Honestly I would have completely skipped this puzzle if not for multiple friends praising it. It seems like a puzzle full of mini-metas, and in any case, surely it should be good, right?

Junky Logic: It's a black box puzzle! (Although right now, it seems like it doesn't work?)

Jigsaw Sudoku: Logic puzzle that I skipped because I was doing something else. But logic puzzle is still logic puzzle, and it should still be fun. Okay, I don't know if these will end up being vanilla Sudokus or what, but at least you can throw them to a solver if you don't like those.

Battle Plans: Another logic puzzle that I also skipped because I was doing something else. There might be some word component involved, but I suspect they are more of a word search, i.e. you don't need to solve clues or anything.

Jigsaw Slitherlink: I saw partial progress pictures while this was being solved, and I think it's delightfully cursed. I didn't get to work on this, but it might be one of the best logic puzzles in the hunt. There might be some math involved.


DISCLAIMER: This is my opinion, my piece. This is what I feel. I'm not claiming they are true or anything, but this is what I believe, and telling me it's wrong will just lead to sour times on both sides. At the same time, these opinions don't reflect the opinion of any other people or any team, not even UE's. Do not take what's said here as some sort of official UE stance or anything; there are many of my UE teammates that disagree with my stance.

This hunt

This hunt was too big and ran for too long.

Honestly, I don't know what made TTBNL think having 238 puzzles was okay. Yes, 50 of them are fish puzzles, but I'd count a fish puzzle to be around half, which gives a number of 213 puzzles. (Some people count fish puzzles for less. Even if you count them at zero, you still have 188 puzzles, larger than average.) This is related to the fact that I think Mystery Hunt is way too large and unsustainable at this pace; see below.

In addition, I feel the rounds aren't cohesive. What was the reason of having 13 Overworld rounds? I don't think there's any specific reason that dictated that number, and the capstone puzzle also didn't need that many. There could be fewer (or more) rounds and everything would have worked out just fine. Normally I wouldn't mind if it wasn't that cohesive, but my complaint was that the hunt could have been shortened easily by pulling a few rounds off.

(For comparison, in 2023, I don't think any of the rounds in Act 3 could have been pulled — and yes, I wish ABCDE:FG had stayed in the hunt. Maybe a round in Act 1 in the museum could have been removed. The problem with 2023 was that individual puzzles were very hard, instead of there being unnecessary rounds.)

That said, I'm also sympathetic for TTBNL. Mystery Hunt is easily the most recognizable hunt, and while TTBNL had some veterans writing, I can also see a lot of excited first-time authors. I suspect they had a hard time saying no, so they accepted a lot of ideas and the scope simply ballooned from there.

I also have to say, although the hunt as a whole wasn't cohesive, the individual puzzles and rounds were pretty nice. Most puzzles were of a fair difficulty and length, so the main reason it's long was because of the sheer number of puzzles, instead of individual puzzles taking a lot of time.

Having a lot of puzzles also means it's more likely for any particular interest to be covered. I don't think we would have gotten an anime meta (Oahu) otherwise. I also found three chess puzzles (A King, Retro, Marchesa), which was a massive number (most years barely had one) — although I'd happily trade one for a puzzle referencing some other work that would make more people happy. (No Pokémon puzzle this year, much sadness.)

Also, I might have been spoiled by the impeccable story of 2023, but I found myself not particularly engaged with the plot this year. We had to escape the Underworld, and then... what? To be fair, I missed the mid-hunt runaround. A teammate summarized the plot as, "destroying Pluto still caused problems in the Overworld and we had to fix them", without much insight of what the problems were. For all I know, we were just doing a road trip across the US somehow. In 2023, I could recognize the story bits better; this might have been helped with the impressive art all over the site, giving all the right vibes; I found it quite lacking this year. Of course, I mainly hunted for the puzzles, so it didn't feel like a big loss for me, but it was still unfortunate.

Tech issues were also unfortunate, but I guess things happen. Galactic and Teammate have a lot of tech people, so they were capable of doing incredible work on the website. Perhaps any hunting team should take some tech people to help, at least to make sure things are stable. I guess this is more a complaint about hunts in general rather than just this year specifically; after all, if we're talking about tech issues, 2016 was probably the worst one in terms of tech (although I also couldn't remember much about that hunt).

But I'm starting to meander off. In short, my complaint is that the hunt was too big and not cohesive enough, I wish they had scoped down, but I appreciated the individual puzzles.

Trend of the hunts

The hunts have been getting too big and running for too long. I'm stealing this chart from Puzzlers Club:

Other than 2017 and 2022, finish times trend upward from 2014-2024

I'm only looking at more recent hunts. I started hunting from 2012, although only actually got more experienced and serious from around 2018. Other than 2013, 2017, and 2022, finish times have been trending upward. Personally, I find 2022 (around 45 hours) to be a sweet spot, and 2018–2021 were a bit long, but still fine. 2023 and 2024 went way too long (and let's not talk about 2013).

If the coin was found late, that meant most teams were late too. Most likely they couldn't even finish, which meant many puzzles ended up not being seen, much less attempted or solved. When you create some creative work, like art, music, and puzzle, you want it to be experienced. Many authors were understandably sad when their puzzles weren't seen, or worse, got berged and skipped entirely. In addition, solvers also got exhausted. A marathon of 72 hours would leave people with no energy, both the solving teams and the running team. If the coin was found earlier, more teams would experience the hunt proper.

I also saw a few comments saying, if the hunt ended so early, it wasn't worth the trip to Boston. Well, first, I think that argument is bullshit, and second, "early" doesn't have to mean the 2017 kind of early. Somewhere like 30–50 hours, where you can comfortably finish the hunt and leave Sunday (either the whole day, or at least the evening) to relax and meet other friends that are gathered with you. (Unless the running team creates an entire virtual world so you can meet and emote your friends there. Galactic was insane. And even then, you'd still want time to hang out.)

The previous paragraph still applies even when you're remote. You have people gathered with you: your whole team, online at the same time. (And for Asian timezone people, a hunt that finishes Sunday afternoon EST means it doesn't actually ruin your Monday morning.)

It also lifts off a lot of burden on the writing team. They don't have to write as many puzzles, which means they can put extra work in other things: polishing the puzzles that are ready, setting up more tech and creative work, and so on.

Of course, it's not easy to resist the temptation of having your work seen by a lot of people, but as I said earlier, your puzzle might not get seen anyway. If the main reason you write is so that people can see it, there are better options to do so — mainly running your own hunt, whether it's more organized like Galactic or more free-form and individual like P.I. HUNT.

I've also started disliking the tradition of Mystery Hunts in the first place. The fact that the winning team is tasked to write the next year doesn't make sense; being able to solve doesn't mean being able to write. And as mentioned in my pre-hunt recap, UE has strong solvers, but we are nowhere ready to write, hence leading to intentionally handicapping ourselves by not using the full suite of help offered to us. This was extremely evident in 2021, where TSBI (which had many of current UE's members, and also not ready to write at that time) and Teammate (a spin-off of Galactic, hence many people were shared in the 2021 writing team) didn't want to win, and the graph of solves over time looked very suspicious.

Some people have argued that writing is "giving back" to the community, and if you don't want to give back, might as well leave the hunt. Ha, I wish. If there was a hunt that was just as grand in scale as MH, I'd gladly drop it; the only reason I still do MH is because it's the largest hunt and likely the most innovative. I do know for sure that, as long as I don't feel comfortable with the culture of MH, that I'm not interested in writing for it.

The tech issues from this year also showed a similar problem. When a team passes the reins of writing to the next one, it seems that little knowledge is being passed over. How did the 2023 site handle complicated requests without problem, and then 2024 had this big issue? The main reason I see is that Teammate is experienced in tech, and TTBNL isn't as experienced; but again, that means there's nobody helping TTBNL with this. Does every new writing team have to make the same mistake?

I feel like I end up rambling and going nowhere with this, so I'll stop it here. In short, hunt too long, hunt tradition doesn't make sense. I'd love to see how — or if — D&M changes anything to make the Hunt more reasonable.