/dev/joe's Experience Constructing MIT Mystery Hunt 2024
It's no secret that for the last year I've been busy helping
create the next MIT Mystery Hunt. But now it's over, and I can
talk about it. Anything that spoils specific puzzles will be in
spoiler blocks. There will be large parts of the article in those
spoiler blocks. Sorry.
This was my fourth time constructing an MIT Mystery Hunt, though
I've also constructed two BAPHLs and a Caltech puzzle hunt, and a
number of one-off puzzles for various reasons. I do feel like I
outdid some of my previous efforts this year, but this article is
not only going to be about my puzzles but about all aspects of the
Hunt I experienced. I also test-solved more than 100 puzzles (129
test-solve sessions, but this includes several canceled sessions
or cases where my session got merged with another). I'm npot going
to write about every one of those, because this would never get
done.
Constructing a Mystery Hunt is a massive endeavor, and it takes
many, many people to pull it off. The wrapup video
gives credit to several of those people but there are many, many
more. I can't thank all those people enough. It is a team effort,
with a massive team; I really don't know how we did it as
Beginners' Luck for the 2010 hunt with so few people. But at a
minimum you can consider this thanks going out to everyone
credited in the wrapup video or in the solution document for any
puzzle.
Also, I should point out that any opinions expressed here are my
own, not my team's. Speaking of which, I have a bunch of other
social media posts from other people (except where I responded in
the AMA). Their opinions are theirs and the posts may contain
spoilers:
If you wrote something and it isn't listed here, it just means I
didn't see it; I'm not excluding anybody on purpose.
TOC
This is more like a novella than a blog post. It's twice as long
as it looks because of the content hidden in spoiler blocks, over
20,000 words! Since you probably won't read it all in one go, here
are links to each main heading and each round of puzzles:
This section includes my comments on the construction effort in
general.
But let's start at the beginning. We picked up a coin in January
2023, after some tough discussions about whether we were ready to
write a hunt. This team (or some semblance to the team, anyway,
since there was a lot of turnover in the years since) wrote a
pretty successful hunt run at Caltech in
2018, but our followup was delayed by various reasons and
then canceled completely when the hunt we were writing about an
epidemic disease that turns people into zombies was interrupted by
a real epidemic disease. But we decided we were ready to give it
another try, and picked up that coin.
First, we had theme selection. There were dozens of theme
submissions, which were whittled down into a short list and then
an even shorter list for a final vote. The vote was contentious
and very close, but we ended up with a mythology theme that
developed into the one you
saw in the opening video (skip to 16:35 for the actual start,
which was delayed because the MIT AV person running lights and
special effects only showed up sharp at noon and needed time to
set up), about the god Pluto disappearing from among the
community of Greco-Roman gods after the planet Pluto is demoted
even further from its status of dwarf planet. I am not going to
mention the other theme ideas we didn't select, since we might go
with one of them if we happen to write another hunt again in the
future. I will say that I didn't really have a horse in this race,
as the idea I had been proposing (for the previous two Mystery
Hunts I had been involved with constructing), The Phantom
Tollbooth, was actually selected for the 2018 Caltech hunt.
(I was also pleased seeing that book earn a spot in Bookspace!)
The construction started off slow, but we got back on track using
a three-region puzzle-writing retreat, a series of weekend
puzzle-writing "virtual retreats," and some weekend test-solving
virtual retreats to get the slug of new puzzles through testing.
We were still behind schedule, but were basically done by the New
Year, and used the last two weeks before Hunt to construct
replacement puzzles for a few puzzles we thought were too long or
problematic. We had a full-Hunt test-solve going in December
(mostly full, anyway, since some puzzles were still being
constructed) but too many of the solvers dropped out, so a few of
us veterans read through the puzzles and solutions those testers
weren't going to reach, pointing out specific puzzles to retest.
In parallel with this we had various creative projects. This
included the art for all the rounds and a few individual puzzles,
and the overworld map. Old maps often drew sea monsters in areas
that weren't well known, so the ones on our map were traditional.
But this one was dubbed Elise the Kraken, and everybody fell in
love with it. if anybody was betting about our mascot (a minor
Mystery Hunt tradition since Zappy the Rat in Bookspace), this is
the one:
Other creative activities included the opening skit and
interactions during the hunt, recorded videos, the coin, staff
shirts, and merchandise (a shirt, pin, and poster). The artist for
the shirt we were selling posted a time-lapse
video of the process of designing the shirt. I was not
signficantly involved with creative tasks, but I didn't completely
skip out on this part of the hunt. I have a cameo in the safety
video about wearing your nametag (the video starts at about 31:50
of the opening video above; my cameo is at 33:13) and another as
Duncan the Swordmaster in the Las Vegas Argonauts video
you could get from the Oracle.
Team Names
It's traditional for teams to have some pretty weird names, but
it felt to me like the names were weirder than usual. Maybe it's
just because of participating in the hint queue and other
behind-the-scenes aspects where I saw many teams' names,
repeatedly throughout hunt. Of course, ever since COVID we've had
many more teams than we used to because there are lots of
all-remote teams who participated for the first time in 2021.
One of the teams had an apparently blank name. One of my
teammates first noted this one, and copying their name into
something that could read it, reported the name was seven
zero-width spaces U+200B. People on the team told us that the name
was supposed to be one zero-width space, but we definitely have
seven on our server. Python tells me the same thing. When I put in
Galactic Trendsetters' name, there are U+FEFF zero width
nonbreaking spaces between the airplanes - but just one between
each two adjacent airplanes. I assume that's to keep them from
combining or something.
There was also a team named canadian geese with three goose
emojis on each side, in imitation of Galactic Trendsetters. And
Team to be named at a later date seemed to be an imitation of our
own name.
Another team's name was a unicode snowman character, apparently
pronounced "unicode snowman." Another all-emoji name was 🪸🐢🪻.
That's coral, turtle, hyacinth. One team's name was just ! and
another's was Interrobang. Yet another team was called [title of
team].
Another team was called now 👇 let's 💢🙏 try this 🅿 garlic
naan. it's paper 🤓📝 thin. 💃💃 [BITE]. 👄 that 😅☝ seasoning is
👌 OUT 😥 OF 🤤 THIS WORLD. 👌. That whole thing was their name. I
mean, it's not as long as Atlas Shrugged, or even the longest
version of our name, but there is a lot going on there. Another
really long name was DAVID ZENG 251 TA BASTION MAIN 🎉NOT A NEET🎉
🚎RB.GY/6QV6Q🚎 🐟RB.GY/55Q9P🐟 42POTATOES CRUCIFIXPANDA ERASMUS
ERAMOOSE TIMEBUFFALO EPOCHBISON YEARDEER 🕓🐂⏳🐃🕰️🦬 TSUKIHI
ARARAGI #1 FAN + friends. I don't know if that is supposed to
encode a puzzle, but the two URLs redirect to puzzles from a CMU
hunt that this David Zeng wrote.
One team was called meow meow. Another team was called 喵喵喵, which
is meow meow meow in Chinese.
A few teams had names that made me think they'd really appreciate
our hunt or specific puzzles within it, including The DONUT within
the DONUT, ℙoNDeterministic, ET Phone In Answer, Destroying All
Nebulae, Resistant to Nixing Pluto, and I'm not a planet either.
Some of those space-themed names might have been chosen after
looking at the space theme on our registration site. However, I'm
not a planet either wasn't new, and for the question where we
asked them to draw an original constellation for us among some
made-up stars, they drew Pluto, with the heart. Not every team's
constellation was going to get in, since we had more of them than
puzzles in the hole in the ceiling, but theirs was accepted
immediately. In a similar vein, The Mathemagicians probably
enjoyed our 2018 Caltech hunt, if they were there for it.
Team Visits
I went out on team visits in costume at one point, as Dionysus.
(Not the Dionysus you all saw on stage during the opening skit! I
was a shorter Dionysus, still with the tie-as-headband, but no
flask.) I visited a bunch of teams. The MIT VISITOR card I saved
from last year was still linked to my account and was reactivated
with a new Tim Ticket this year. Unlike last year, when I barely
used the card, I had plenty of opportunities to try it this year.
A few times it seemed not to work on doors where it should, even
though it gave me a green light, but my experience at Building 46
made me think some of those doors were just broken. The entrance
there facing Vassar Street features a normal door, a revolving
door, and another normal door. Only the normal door farthest
from the card reader opened for the card.
One team we visited was The Girlz and otherz and GPTz. This team
seemed to consist of six Chinese "girlz," perhaps college age,
with an older Chinese man sitting at the back of the room, not
participating. Their supervisor, guardian, or something. Their
English wasn't perfect, and was strongly accented, but they were
able to hold a conversation with us with only a few instances of
one of us not understanding the other. There were a few of these
all-Chinese teams, some of them only playing from China, who were
among our biggest hint-requesters, often citing language and
cultural issues they had getting through the puzzles. I applaud
them for trying, though they are far outside our target audience
and the Mystery Hunt is not written for teams like them. (I was in
a similar position about 15 years ago when I was among the
pioneers in trying to do Austrlian puzzle hunts from the other
side of the world. At least we spoke mostly the same language, but
we struggled with the cultural references at times. One puzzle was
about the card game 500, which despite being developed in the
United States, is now almost completely unknown outside
Australia.) We heard from one of these Chinese teams (I didn't get
the comment directly and I don't know which team) that they
weren't intending to abuse the hint system, but they had one
member they couldn't control who was just constantly asking for
hints any time the button was available. Team behavior management
is another issue entirely and one that fully English-speaking
teams also encounter.
I also visited Control Group, who had a full classroom, about
half of them wearing cat ears. I'm not sure I've encountered them
before in person, so I don't know if it's usual that half of them
are pretending to have taken an experimental drug that turns them
into cats. (They had headbands with cat ears.)
Finally, I went out on the inaugural run of the interaction after
solving the Hydra. It was supposed to be result of clearing a
server of a computer virus which generated popups of a snake named
snek, spawning new copies the way the Hydra spawns heads.
(Incidentally, after I saw a mockup of the Hydra round page full
of snek popups, I posted a link to the Vi Hart
snake snake snake snake snake video, which one of my other
teammates was coincidentally looking for.) For this interaction, I
wasn't portraying one of the gods. Instead, I was the grateful
server admin. In the interaction, we had several people get in a
circle, put both their hands in the middle, and clasp hands with
two other people. Then we asked them to untangle themselves
without letting go of any hands.
The Server Troubles
At the start of the hunt, we experienced severe server troubles.
I don't understand the full technical detail, but the setup
involved some middleware between the web server and the database
which was apparently supposed to detect when either of them was in
need of a restart. My impression is that there were probably
memory leaks which restarts cured. But in this case, the
middleware itself was getting bogged down, so it didn't do its
job. Post-hunt discussions seemed to lay the blame on websockets,
but I don't know what it is about websockets that we were supposed
to do differently.
What we eventually did was strip out this layer, so the server
talked directly to the database. This made it "fail quickly" in
the words of one of our tech gurus. Instead of gradually slowing
down to uselessness, either the server responded quickly, or it
generated a 500 error quickly, at which point we ran a command to
restart it. That worked, but we had essentially replaced a broken
piece of software with humans, to restart another broken piece of
software, and we added more humans later to keep it going 24 hours
a day. Once hints were going, we had people in HQ hitting the
server pretty frequently to grab new hint requests, and hinters
either included somebody given the rights to "kick" the server or
they'd call out for someone to do so. Ironically, the hint queue
page didn't "fail quickly" but it usually either responded
instantaneously or hung, so we were still able to use that as an
indicator.
While my teammates were trying to figure that out, we worked on a
plan B. The basic idea was to open a Google form to submit
answers, and do answer callbacks, after giving them all a copy of
Google doc versions of some puzzles. But it didn't work to do this
for the first round, mainly because we'd have to find and email
the meta-related images to each team that they were supposed to
get when solving the feeder puzzles. The second round contained an
interactive puzzle that there was no easy way to send out, so we
sent out the third round.
We had Google Doc versions of most puzzles because of
autopostprod. That's a script that automatically postprods simple
puzzles made of standard content if they are provided in Google
Docs, turning them into the web pages that you see. But we had to
quickly find and check the Google Docs for all these puzzles, to
make sure there weren't last-minute changes. Our factchecking crew
handled this job in about an hour. You can thank them for having
these puzzles to work on while we worked out the server issues.
Ironically, it seemed like we got the server to the form it
stayed in throughout the hunt about half an hour after you guys
got these puzzles. However, because teams didn't actually have
this round unlocked on the server, we kept the Google form and
answer callbacks open for several hours, until we opened the round
on the server for all teams. I was on the last shift of
responders, and it was fun to hear some teams that still
remembered their funny ways of answering the phone, such as ET
Phone In Answer answering the phone with the answer they last
submitted in place of "Answer" in their name, and Wafflehaüs
asking to take our order.
Opening up more puzzles also had the side effect of speeding up
the hunt slightly, but since everybody had lost a couple hours at
the start, we felt that was fine.
The Puzzles
In this section, I describe individual puzzles roughly in the
order you might have seen them during Hunt, because reading them
in the order things were constructed would be too confusing. Also,
this lets me make sure I don't talk about puzzles that weren't
actually released in here, which would be even more confusing.
For all puzzle links below, first go to https://mythstoryhunt.world/
and click Public Access to log in. After that, you should
be able to navigate directly to the puzzles via their links below.
Queen Elizabeth II had corgis, but it was new to me to learn
they were all related to one another. But once you made that
discovery, this was easy to work out; a good first-round puzzle.
This was the most contentious meta during meta development, and
it almost didn't make it. The version I tested was rather
different.
At that time, the diagrams you got when solving the feeder
puzzles had colored lines, matching the coloration of the
members of the trios, with color mixing where the same line was
used for two of them. It had the same aspect of having to match
letters to locations in the diagram. Rather than the judge
scores, there were missing lines, and it seemed obvious to me
that we should join up the missing lines of each color to spell
a three word answer. Except that wasn't it; we were actually
supposed to keep the puzzles in the order given (reinforced in
the final version by forcing the first letters of the puzzles to
be A-F) and read the missing letters in RGB order of the lines
missing from each puzzle's diagram.
There were several other versions, both before and after that
version, which used different mechanisms, some different feeder
answers, and several different final answers as well. I think
there was a serious too-many-cooks thing going on here, with too
many different concepts being proposed for what was supposed to
be an easier, early-hunt meta, and combinations of them getting
put together which didn't work well or made for unintuitive
steps as in the one I faced. If you were wondering why there was
a color theme among the trios for the answers which wasn't used
later, it was because using it made solvers want to use it more,
in a deeper way that wasn't intended.
This was another failed of my test solves. Really, the failed
ones were a tiny portion of the testing sessions I was in, but
they included two out of the first three I saw as interesting to
mention in the order they were presented during Hunt.
We didn't have the flavor text clue that was in the final
version, and my co-solver and I had never heard of the game and
didn't recognize the bridge image. With the same sort of grid (a
different but still irregular grid shape, with differently
arranged flowers, and some different flowers) as the final
puzzle, we ended up down a deep rabbit hole trying to make this
into a cryptic Nurikabe which didn't lead to any quick
contradictions, but also didn't lead to enough certainties to
make it solvable. This isn't even the biggest BE NOISY that I
encountered during this year's testing (See Obelisks of Sorrel
Mountain in Oahu for that). But I bet the reason you have
flowers adjacent to each other is to close that rabbit hole.
Somehow my group testsolved this one without ever paying any
attention to the name, which we would just have noticed as
a sign with letters knocked out,
not being familiar with the meme. This is the problem with
meme-based puzzles: They are usually not as widely known as
authors wanting to use them in puzzles think they are. I admit
that giving the meme explicitly as the title of the puzzle does
make it a lot easier to find. And in this case it just didn't
matter, as the puzzle was solvable without understanding there was
a meme in the first place.
This is the first puzzle in this review which I didn't test
solve, but saw during the hunt and decided I had to try it myself
afterward. The title alone is worth it. The logic is a difficult
slog, though.
I was assigned to discussion-edit four different chess ideas
various authors submitted early in hunt construction. The editors
thought it was too many for the hunt, and were hoping I could pare
them down. The one I liked most ran into an unexpected technical
snag, so got booted out. The other three all needed some level of
guidance and ultimately all got in.
This was my second choice, the only one I thought could be easy
enough to be an underworld puzzle.
It was a simply fairy chess puzzle, where solvers are going
to get only indirect clues (and plenty of examples) to figure
out how a Cerberus piece moves.
The puzzle stayed true to its goal, but took a long time to
develop once the authors latched onto this idea of cluing a
metapuzzle using FEN notation, one row per puzzle, but had
difficulty coming up with a position that did what they needed
with FEN that could be constructed using characters from
algebraic notation for chess moves.
I didn't actually see this one in testing, but because it
involves the badges we were handing out to all the teams, it was
talked about a lot in the lead-up to hunt, in part because we were
trying to figure out how many of those badges we needed to print.
This was more than just adding up all the on-site registration
numbers because we were giving small teams a set of 8 badges to
ensure they could complete the puzzle, and wanted some spares in
case some badges were lost or ruined. We had people constantly
making references to badgers,
mushrooms, and snakes and not needing
stinking badges.
This was a case where I was called in to get a test-solve group
unstuck.
They had already figured out the movies, and even the
mechanism, but they had some bad matching of the items to the
categories indicated by the movies. Even after I sorted them
out, a couple of the matches felt weak, but I didn't really have
any better ideas. My feedback included capitalization fixes so
that the two-word phrases consistently had lower case on the
second word, except for the proper name. I gave out some hints
during hunt that amounted to data checks of this puzzle to help
similarly stuck teams.
A pretty quick solve, but it took a bit to get the idea
that they were consonantcies. I think we got started with Rex
Stout matching part of Our Ox Stoat with two letters changed and
as we made other near-matches figured out the consonantcy
aspect.
This was one of the puzzles I found the most fun in test-solving
(keeping in mind that I missed many easy-but-fun ones that just
didn't need my help to get them tested) and I highly recommend you
try this one before spoiling yourself. 2 or 3 people with voice
chat to brainstorm what's going on is recommended.
We got started in the middle, with Deep Blue Sea a solid
starting point to figure out that the 7 extra squares beyond the
emoji bank needed to be filled in with letters. We had HIC, and
I found a lot of 7-letter words with those letters. Then we
finally figured out the one about ET was supposed to be
"ET phone home," and then I didn't need a word list and called
in the answer.
I had the idea from the flavor text that this one was about
the MIT Blackjack Team even before we looked up the title and
found it was the title of a book about the team. Reading about
the team led us to the knowledge that they used a set of
numerical signal words, and once we found those it was
practically done.
This was a puzzle that came out of brainstorming at the New York
retreat which I implemented. You probably missed it, since it was
an easy one that whoever opened it first probably figured out
quickly, so go take a look.
This was my only other puzzle in this round. I apologize if it
ended up too long and difficult for the round. The idea was really
meant for an early overworld round, but the editors didn't have
any good answers for me. On top of this, the answer was hard to
locate; I and the editors both missed that the answer isn't in a
lot of word lists and the anagram I tried to clue is nontrivial.
The puzzle was still nerfed from what it might have been in a
later round. An outline for such a puzzle included six of each
variable instead of five, and made it a puzzle not just to match
the total emissions scores to the companies, but to determine them
from even more clues.
This one starts out with an ordinary crossword puzzle, so if
that's your thing, it's a good one to try.
I enjoyed it during test-solving. During hunt I hinted a
number of teams who were not looking at it the right way, and
thinking of words like SLANT and UNION, telling them to find
longer words within the math theme.
Poor Seth, though, had to redo the whole puzzle because he had
misspelled one of the subjects TRIGOMETRY in the version I
tested (and somehow the fact-checker missed this).
I enjoyed this one. (An apology for some solvers is in the
spoiler.)
It was called Making Connections when I started testing it,
and while I got the Connect Four connection pretty quickly, we
didn't figure out the mechanism to change letters in the clue
answers to red or black until the title got changed to what you
saw.
A lot of people seemed to like this one, but we got feedback
from a couple people saying they got the color replacements, but
didn't think it could be Connect Four because that game has been
played with red and yellow pieces for a long time. Some
research shows the first Milton Bradley versions with red and
yellow pieces debuted in some countries as early as 2007 and in
the US in 2009. Oops! Definite generation gap problem. I played
it with red and black and the authors are in that generation as
well. Same for the hint writers. Sorry, Gen Z red-and-yellow
Connect Four players.
This was another chess puzzle I discussion-edited.
Andrey Gorlin linked to a set of chess rebus puzzles he had
seen elsewhere (basically cryptograms where the identities of
chess pieces were hidden and you had to do retrograde analysis
to figure out which piece was which, as makes up part of this
puzzle). And after skimming through the solutions, I saw that a
massive variance in difficulty was possible. You could make one
of these puzzle solvable in a few minutes, and you could make
one that took hours and hours to solve and several pages just to
explain why the solution was what it was. So my main advice was
to keep the puzzles on the easier side, because a Mystery Hunt
puzzle would likely involve solving several chess puzzles to
give solvers enough fodder to extract an answer from.
He took the puzzle into a domain I was completely unfamiliar
with, so I mostly just observed once the actual construction
began. He needed to nerf things at times, but delivered a
solvable puzzle in the end.
I was in a long, difficult test-solve session for this one. Of
course, almost everything I can say about it is spoily.
Understanding the rule roll frequencies was simple enough,
once you collect the data. We had a lot of trouble figuring out
certain groups for assembling the die, and a few of the numbered
spaces. More of a problem was matching the D&D rules.
Several clues were changed as a result of trivia we unearthed,
and we assumed we'd use AD&D 1, 2, 3, 4, 5 before the list
of editions above the rule roller was added.
Finally, we had trouble seeing some of the letters we were
supposed to be forming. At least one of those was changed to be
a clearer letter as well.
Haha, it is pretty routine, in that it has a mild gimmick that is
easy to figure out. But it's also huge. If you love cryptics, this
is a good one to keep you busy for a while, and if you hate
cryptics, go find another puzzle. This was the second puzzle
(after the Romans one) that I went back to solve after hunt was
over - not an entirely clean solve, since I'd seen the answers
once, but it was months earliand I didn't solve most of the clues
myself in testing and didn't remember much.
I joined a test-solve team that had already solved all but a few
of the cryptic clues, so I missed that particular joy.
They had the messages from the extra letters too, and knew
that The Naughty Nineties had Who's on First in it. They
were struggling to figure out what to do with that. I was the
one who found and pointed out to them the giant diamond filled
with Who's on First characters.
This was one of the earliest non-meta puzzles test-solved, and I
was among the early testers.
The nonogram wasn't too difficult. But then what? Knowing
there were Minecraft symbols in it, but not actually having
played Minecraft, my first thought was that the ^ and v symbols
meant to go up or down a level. Later, with more insight as to
what Minecraft patterns looked like and in particular the
placement of the lamp cells, we figured out they were origami
fold lines, and we got the real 3D structure.
Then we had to try to figure out how the redstone logic worked.
This was complicated by the fact that the version we started
with had peashooters from Plants vs. Zombies instead of the
redstone repeaters. I am not sure why that happened; maybe the
author was learning more Minecraft along with us.
Eventually we did figure out how the redstone circuit was
supposed to work and solved the puzzle properly, with some
hinting that included linking us to a redstone simulator.
As a Strongbad fan from back in the day, this video brought back
memories. You don't actually need any knowledge from anything on
homestarrunner.com to solve it, though, and it's a fun little
puzzle, if your idea of fun allows a little comical cartoon death.
It looks like there's a story I have told people but failed to
blog after the 2016 hunt, so first, a digression for that story.
There was a puzzle called World's
Longest Diagramless. The gist of the story is that as a
test-solver for this puzzle, they made me write a program to cheat
at crossword puzzles. Details in the spoiler block.
The puzzle is posed as a giant diagramless crossword for
which even the clue numbers are withheld. At the top and bottom
it behaves like a normal diagramless. After a bit, though, it
turns into a list of alphabetical words. Not knowing what we
were looking for, I downloaded a copy of Matt Ginsberg's
crossword clue database, wrote a little Python to pull the clues
and answers out of the simple null-delimited database format,
and then wrote more code to look up the answer for every clue
that appears with exactly one answer in the database as an
assumed correct answer. Of course, this didn't catch all the
words, but it got enough. And we had enough data to figure out
what we were looking for.
It turns out it's related to "Yakko Warner Sings All the Words
in the English Language," one of several Yakko Warner music
videos from the Animaniacs cartoon. The first and last
alphabetical segments are the words he sings. In between, it
just had all the otherwise unused words from some word list that
there were existing crossword clues for. Of course, the video
skips over most of the words, but there is a segment in the Ls
that he also sings, and those words are also given together near
where they belong alphabetically, and the answer was hidden as
an acrostic in the clues at that point.
Now, back to the Git puzzle. This was this year's puzzle that
made me learn to cheat in a new way. The puzzle claims to be a Git
respository in which the commits have gotten into the wrong order.
Now before we proceed, I should explain my background. I know a
little about Git, but I'm no expert. I've done basic operations.
Updated my local repo and sent pull requests. That sort of thing.
Now on to the spoilery content.
At first glance, I thought it wanted me to learn how to use
git rebase to patch the commits back into the right order. But
when I sat down to solve the puzzle for real, I said to myself,
why not just make a new repo and shove all the commits into it
in the right order? I just needed to understand what needed to
go in.
git log -p showed me the history of the repo in a
readable form, including the commit hash. And they gave us the
hash of the first commit so I could verify I was doing it right.
Setting the file contents was easy. I figured out how to set the
user name and email, and how to set the date within the git
commit command. After it didn't work, I found a document
(much like the one the solution links to) which explains there
is a second committer date that also also goes into the hash.
This was more difficult to set, and the page I found describing
it suggested it couldn't be done with a command line argument
and had to go via an environment variable. In other words,
probably something people don't ordinarily do.
Once I confirmed I had the first hash right, I programmatically
did all the rest. I'm sure this was great confirmation to the
author and editors that somebody who only knows a little about
git and looks up the rest can figure it out. And I hope those of
you who tried this puzzle during hunt figured it out too. And
that's how I learned a new skill: How to forge a completely fake
git history. I hope I never have to do that again.
Once upon a time there was a team called This Space Intentionally
Left Blank. For one of the BAPHLs my team ran, we actually had two
teams with this name, the traditional one and a new team who
didn't know their name was already taken. And we didn't have a
system that forced the names to be unique. The new team had
entered in the beginner division that gets extra hints, so we used
the division names to distinguish them.
I'm sure they (the first Blank team) are still around somewhere,
though I'm not sure which team they are playing on now. This year
we had a team whose name was seven zero-width spaces. Or just one.
Not that it makes much difference.
Oh yeah, this puzzle. The puzzle tells you there's no puzzle
content in the page, and it's all in a Google doc... which also
looks blank. It's a fun concept which I didn't get to test-solve.
I didn't test-solve this puzzle but experienced it during hunt
through the interaction (which reveals the answer you get from
solving the given puzzle, so is in a spoiler block).
The puzzle solves to the answer HOLD BILINGUAL DIALOG. When
they submitted their request, we called teams and spoke to them
in some other language, which was commonly Spanish, Chinese, or
Japanese. Galactic Trendsetters got a call in Puflantu!
This was a fun one. The puzzle changed a bit from when I
test-solved it; there was another slide, some different ones, and
a different final answer.
My solving partner got that this was about esolangs before I
even had a chance to look at it.
We didn't have the punctuation in our output, and we wanted to
apply what were originally triples of numbers as epigram, word,
and letter indices, as some book codes do, but it didn't work.
It took longer to figure out we wanted to extract more than one
letter per epigram.
The version we got, rather than just looking like slides from a
Powerpoint document, was actually in a Powerpoint document,
and since we got stuck, we were considering whether there was
another esolang to find. In wondering whether Powerpoint itself
could host an esolang, we found this,
which we were easily able to rule out, since our Powerpoint did
not have these features, but it was funny nevertheless.
Another weird bit is that when you look for the "Epigrams
on Programming," Wikipedia links to two archived copies of
it: A PDF from the original publication, and a copy of the
author's web page in the Wayback Machine. The web page version
misspells the word Optimization as Optimiziation in epigram 21,
and of course the version of the puzzle we received used that
one and we had to guess whether to treat the misspelling in the
source material as canon. (It was easier to decide to ignore it
when we saw the other copy without it, but we also advised just
not using this epigram, which they did.)
I proposed this meta a bit later on, after several metas had
already passed testing and were having feeders written, but it
ended up being one of the first rounds solvers saw once they
arrived in the overworld. Team leadership put out a call for metas
with unorthodox structure, worried that a feature which has made a
lot of Mystery Hunts (especially recent ones) memorable was
missing from our hunt. This was my response. It's clear that a
whole lot of you loved it. The Text Adventure seems to have been
the most loved, and that wasn't mine, but maybe I get some partial
credit for commissioning a teammate to write a text adventure for
me with certain details and letting him go wild on the rest!
At least a few people were stunned by the possibility that this
could be constructed, so I've made a lengthy writeup about how it
was done. I wanted to say this at wrapup,
but they only gave me two minutes to speak, so I couldn't do it
justice. (Those two minutes start at about 14:50.)
The inspiration was the Szilassi
Polyhedron. This bizarre shape is a polyhedron with 7
faces, each bordering all the other faces. The seeming
impossiblity is made possible by the shape having a hole through
it, and most faces being concave.
I used the shape for a puzzle in the 2018 Caltech hunt, but
didn't really figure out how to make good use of it then, and
ended up just making a lame maze. Once you cut out and piece
together the shape, and then solve the maze, it forms a letter
on each face.
This time I knew how I wanted to do it: Seven puzzles, each with
six answers. None of the puzzles are independently solvable, and
the answers are shared by pairs of puzzles which contribute
information toward each answer, so there are actually 21
different answers. The scale was reasonable, on par with other
multi-answer rounds like the Zelda round
(9 puzzles with 3 answers each = 27) and the Orbital
Nexus round (8 puzzles, each with 4 versions that switch
with the phases of the moon = 32). It also was a bit like the Reverse
Dimension (18 puzzles which had to be paired up to make 9
answers). And yet, it was significantly different from all of
those.
Since I was putting a puzzle on each face, and an answer on each
edge, the natural extraction was to put a letter on each vertex,
and I chose the mechanism that the three answers meeting there
would have only one letter that appeared in all three. This gave
me a good flexibility to come up with answers, and it let me
impose the second constraint that every two answers meeting at a
vertex should have at least two letters in common, so there is
still a reason for solvers to keep solving even when they have
many answers already.
Another issue that came up at this point was how to read those
letters. The problem with Szilassi is there is no obvious order
to read the letters. There is more than one Hamiltonian; some
site told me there are 24 Hamiltonian loops (each with 14
starting points and two directions to try to read the answer),
and more orders if you don't force it to be a loop. But the
seven-colored torus has equivalent topology, and has the
advantage that a common representation of it has all the
vertices in a circle. Googling seven-colored torus gives you
multiple images like this, including somebody selling 7-colored
torus pillows in this pattern. By doing this, there was one
obvious loop to read the letters in. Solvers had to figure out
the starting point and direction still, but that's a reasonable
puzzle. And I decided I was going to give solvers the plain side
of the torus (with just seven diagonal bands) and suggest to
solvers to turn the torus over and imagine what it must look
like on the other side, if we could get it through testing that
way.
But there was a gap to fill still, a Grand Canyon-sized gap:
The gap was the difference between saying every pair
of puzzles is going to combine to generate an answer and writing
seven puzzles that work together so intricately. So in addition
to proposing this mechanism and a set of answers, I proposed a
full set of puzzles for the round and the way the interactions
between them were going to work.
First was the Blanks puzzle. The point of this puzzle was to
help solvers understand these weren't regular puzzles right up
front by giving a puzzle which clearly didn't have enough
information. There was just a set of blanks with 5 of them
marked. Each puzzle generates or has hidden within it one phrase
with words of the right lengths to fit on the blanks and extract
a five-letter answer.
Next up, I chose three grid puzzles of the same size, the common
15x15 size for crossword grids. One of these would in fact be a
crossword, one a word search, and one an Akari. The lights of
the Akari overlay on the other two puzzles to extract clue
phrases, and leftover letters in the word search would tell how
to get an answer when overlaying it on the crossword. At this
point, I thought the Akari wouldn't have any letters, so the
Blanks phrase was going to be given in its flavor text. The
crossword would have a clue, and the word search could have more
of the unused letters.
The fifth puzzle I added was going to be a Matchmaker. I adopted
this name from one Mike Selinker used in Puzzlecraft to
describe this puzzle when I put it into my index. It's the one
with clues with dots next to them which you have to pair up,
draw lines between them, and look at what the lines cross, or
what they don't cross. By making a really overloaded Matchmaker,
I could make it extract six different phrases, one to clue an
answer with each other puzzle. It could use one of the crossword
answers, one of the word search words, and provide a phrase to
put on the blanks. I wasn't quite sure how it was going to work
with the Akari.
Nathan Fung suggested a technical puzzle, for which I chose a
chemistry puzzle, and an interactive one, for which I chose one
of his suggestions, a text adventure. I had never done it, but I
knew we had people who had written text adventures in Inform 7,
and figured I'd farm it out to one of them to construct. The
chemistry puzzle could have clues pointing to specific
chemicals, and a phrase from marked blanks in the names of
missing chemicals to point to another puzzle, and the text
adventure could have clues coming in that give easter-egg-like
hidden commands and give out phrases for other puzzles when you
complete whatever activities the adventure involves.
As I started looking for a metapuzzle answer, I noticed some other
constraints the mechanism I had proposed put on me:
Certain combinations of positions could not have the same
letter, because making the answers leading to those letters all
have one letter in common would force another letter to also
have that letter in all its answers, even if I didn't want it
there. I developed a list of such combinations and wrote a
program to help me test answers for problems of this sort.
Eventually I decided on an answer and got editorial approval for
it.
I actually wrote the Blanks puzzle first, before the meta was
even tested, to ensure I had a set of answer words for Blanks
which could be extracted from some sort of plausible phrases,
with one crossword clue, while the others would be extracted
from the solutions or flavor text of puzzles and could be
somewhat contrived. I placed the puzzles in the grid as I chose
answers, putting a bunch of chemistry-related answers around one
region for that puzzle, and then working to meet the constraints
I had set for myself in the other answers while choosing
sensible and interesting answers.
Testing was a bit awkward, because we had to figure out what
information to give testers. It was during this phase that we
decided on the idea that solving all answers for a puzzle was
going to give solvers the color in the torus that represented
that puzzle. Finally, we got two clean test-solves... And then
an editor wanted to check that the answer wasn't too
brute-forceable.
The answer I had at this point was DISHARMONIZING, a pun on DIS,
another name of Pluto, saying that he harmonized the groups and
kept them from fighting. But it was also a real word and in word
lists, and brute force tests on 11 answers (everything from two
puzzles) and 10 answers (some other combination) showed just two
answers for the 11, and a handful of answers from which the real
one could reasonably be picked for the 10. My editors were too
worried that someone was going to plug an expression into
Nutrimatic, or more likely 28 of them for all possible starting
points and directions, and get the meta answer with only 6 or 7
answers.
One of my teammates came up with another answer that was deemed
both sufficiently punny and sufficiently non-nutrimatic-able,
and I redid the whole thing, including a completely new set of
phrases for the Blanks puzzle, and more clean solves.
Finally, we were ready to start constructing. I had planned the
order for construction as well: Blanks was already done, Akari
was next (which I got onto immediately and done in a couple
days), Crossword next (which Craig handled), and Word Search
next (me again). Text Adventure and Chemistry were next, and
this happened to be when we had scheduled the in-person retreat.
Craig found me Linus Hamilton, who had a personal goal of making
sure there was a text adventure in this hunt. He said writing it
himself was an acceptable way of meeting the goal. I had planned
to write the chemistry myself, but I met Alina Khankin during
the retreat, who was eager to write a chemistry puzzle, so I
gave her this one and put her in touch with Linus to coordinate
the message linking those two puzzles.
And finally, I wrote the matchmaker to give all the messages
left over needing to point into all the other puzzles. The
erratum for this puzzle came about because Linus gave me the
message using UP, but implemented the command without that word.
I am not sure how it got through two full-round regular test
solves, fact checking, postprod checking, etc. without anybody
picking up on the problem.
This was Charles' idea, but he asked me to build the map, knowing
getting the scale and detail right ws a problem. He designed the
concept and the extraction cluephrase, which was revised due to
testing, so I had to do the map twice. Writing a good cluephrase
was a bit of a challenge, because this data set and mechanism
rather limited the possible phrases.
This puzzle was not my idea, but I necessarily got involved.
It's pretty much in the same form as I first saw it, with
only minor refinements. They just needed me to not make any
updates to my index before hunt, save for adding this
puzzle to the index to make the first diagram true. So I had the
update staged and ready to post weeks in advance, and posted it
Thursday night. If some of you saw it in advance, good for you!
It probably didn't help much, since when you got the puzzle
content it linked back to the index.
This puzzle involved information that will change over time. An
archived version of the information at the time of the hunt can be
found in the next spoiler box.
The keyword list and the authors list. Specific pages
you can reach from links from those pages are also preserved;
everything else will error out.
In the original puzzle, there were only a handful of picots.
We were supposed to recognize letters like in the final puzzle,
but we were supposed to shift those letters by the number of
picots in each one, a step which was entirely unclued. Due to
nobody figuring this out and on top of that having trouble
recognizing the letters, the design was updated to draw the
answer, with letters that held together better due to not having
any restrictions on picots.
The first meta I proposed, and one of the first ones to get
through testing, was one based on...
the Mexican bingo-like game Lotería. Though I am not
Hispanic, I grew up in south Texas and experienced many elements
of Hispanic and especially Mexican culture, and this was a
subject I had long considered a candidate for a puzzle but
hadn't ever really developed.
It wasn't a smooth process; my first proposal had problems, and
somehow I ended up with three different editors making
contradictory suggestions about how to fix the meta. This was a
pain; it meant I wasn't going to please everybody, but I just
picked ideas one at a time and tried them. We went through a
bunch of different ideas on how you could "play Lotería" on the
board (which was going to have pictures of the 16 casinos laid
out like a board on the round page, filling in the answers at
the bottom of the images like the names on a Lotería card when
the puzzles were solved) but we just couldn't come up with
anything that worked. Just picking a few lines to play in meant
some answers wouldn't get used at all, which was unsatisfying,
and trying to play in all the lines (which is more than you
think; an
animation here shows all the traditional ways to win) left
it so overconstrained we couldn't make anything work. Eventually
I picked one of the other ideas another editor had suggested and
went with that once we got something decent out of it.
A second round of editorial upheaval this puzzle went through
was related to a few images some people might consider
culturally insensitive by today's standards in the 19th century
Lotería card set that has become standard. I was aware of those
(they include a drunkard and a black man) and I had already
marked them as not to be used in any way in the puzzle, but
teammates were worried about even potentially subjecting solvers
to these images by making them search for the game. Someone
suggested alternate, modern Lotería games which have been
published, but there were too many different ones, some without
good sources to identify all the cards. Instead, we identified
what Google had done when they had used the game in a Google
Doodle, which was to keep the traditional cards but replace all
potentially objectionable cards with safe images, including some
Mexican cultural references they thought should have been in the
game, such as the Mexican hairless dog. We clued the game that
way, to at least provide a way for solvers to get in which
wouldn't show those images. The final version of the meta used
neither the cards Google removed nor their replacements, letting
it work with whichever list you used.
It's hard to say much about this puzzle that isn't spoily, but
I'll say it's a very deep puzzle and not the sort of thing to
attempt over your coffee break.
My testing group identified it was about Sleep No More
pretty quickly. There are lots of hints for that, and we found
even more later on, such as the fact that image with the icons
above the door at the start of the puzzle depicts the actual
door to the New York location for Sleep No More, to the
extent of locating it on Street View and showing that the star
shape and pineapple are there.
Before I go any further, I want to point out that solving this
puzzle may subject you to spoilers for Sleep No More.
When we tested this we thought that wasn't much of a concern
because its run was supposed to be ending shortly after the
hunt. Now I see its been extended to the end of March. But I'm
not spoiling it in any major way here.
The first level of solving this puzzle was to identify the
scenes. There's kind of a hush-hush attitude toward taking about
Sleep No More on the Internet, but there are a couple
sites with a lot of details. This both funneled us into those
sites to get the info and let us know the info we need from the
play isn't too in-depth because what's available is limited. In
any case, we figured out the scenes and the main character
identified as SOMEONE in each description.
It took longer for us to figure out what to do next, and we
explored every possible aspect of the play we could get from the
Internet. Eventually we got updates to the flavor text that made
the idea of looking for hidden items clearer to get us to the
answer.
This was a puzzle I came up with during the NY retreat.
Daniel Kramarsky was trying to develop some other kind of
puzzle based on 3x3 grids, but his puzzle concept was too loose,
with each grid working differently, and I worked with him a
while before abandoning the idea. But the 3x3 concept made me
think of the D&D-style alignment grids for all sorts of
things not related to D&D which were a popular meme in 2023,
including one for crossword grid symmetry. This led to the
puzzle you got.
Before I ever managed to get the grid filled, I figured out
the rules were equivalent to an allowable sequence and did the
right searches to find that term. So I was confused for a bit
when the grid extracted to ALLOWABLE SEQUENCE. But not long, as
we were given the diagram of the square and I realized this
meant to fill it in.
This was a grueling long test-solve due to the number of false
matches we kept finding. A bunch of those clues got rewritten and
rewritten again. The casting costs weren't originally given.
There were false matches to irrelevant Magic cards, and false
matches to unintended Iliad scenes. The casting costs
helped us eliminate some of the false matches to Magic cards
which are hard to avoid simply because of how many cards there
are today. A bunch of cards with one- or two-word names showed
up by accident. Different cluing, and getting enough of them
correct to understand that we were supposed to have all the
clues in the first round of the tournament reference Iliad
scenes before any of the ones in the next round, and so on,
helped us nail down the Iliad references.
Finally, there was a discrepancy between playing the cards to
help that tournament participant (possibly on his opponent) and
playing them on that participant. They hadn't intended to have
Disenchant in there at all, but they had another
enchantment-killing card accidentally mentioned on (for one of
the ways we were wrongly lining up tournament contenders in one
version of the puzzle) the opponent of the one enchantment among
the Theriad cards, and this made me think "sure, that's
how we deal with a non-creature being in the tournament" when it
was supposed to just be an automatic loss for that card. In the
various revisions we ended up with Disenchant being in that
card's clue to help point out you are supposed to cast it on the
compeititor whose clue had the card.
Originally this puzzle was supposed to have some Chinese writing
on the right side, but somebody thought using the Chinese language
that way was culturally insensitive, so when I was called in to
rescue this puzzle, I drew the thing that actually appears on the
right.
My testing group never figured out one key aspect of this puzzle.
The part where the routes are ones from famous bicycle races.
We found most of them the hard way, by searching for routes on
the site of the right length and in the right country, and
anagrammed extracted letters within each section to figure out
the answer.
I joined a test-solve already in progress to get them unstuck.
This spoiler comes with a bit of an ICK warning too.
They had figured out the answers that fit the full quiz bowl
questions, and some bonus answers, but that's it. At some point
after I joined, one of them noted the three-sentence format of
each question, and we started considering them separately. It
was when I got down to the sentence that reads merely "This
figure fathered Aphrodite" that I got my aha. Aphrodite has two
origin stories! A quick lookup confirmed: Zeus fathered
Aphrodite by Dione per The Iliad, and the other more
gruesome origin is that, when Cronus overthrew Uranus, he
castrated him and threw his genitals into the sea. Aphrodite
sprang forth from them and by some accounts rode a shell to the
shore, as depicted in Botticelli's The Birth of Venus. A
little further checking confirmed that Uranus is considered her
father in this myth. Good thing we didn't need to know her
mother!
We went on, gradually beginning to find non-Greek myths fitting
the clues, and realized the flags gave us the locations the
myths are from. I also contributed that the answers to each
bonus should be sorted alphabetically, confirmed by getting one
selected bonus answer to match each tossup answer when we did it
this way. Finding some of the more obscure myths was the hardest
part, but the flags guided us into at least looking at the
correct cultures. It was after I found Teshub, the Hurrian storm
god who was tricked into eating a rock, that we had enough
letters to read out the clue phrase.
I got called into a test-solve on this to add my expertise.
They had figured out the art was Grateful Dead-related, and
wanted somebody who knew their music. I am not the biggest
Deadhead by any means, but I was who they found. After listening
to all the music samples, I told them these aren't actually
specific Grateful Dead songs; it's when they are jamming, for
example at the bridge of a song in concert. It sounds like Dead
music because it uses musical phrases from their songs, but
sometimes they are mixed together. Where I was able to identify
them, I pointed out examples. And I likewise told them I had no
idea where I'd find them, and even if I had the recordings of
those concerts, I didn't have the time to go listen through them
all for matches. And I was on so many other test-solves that I
really meant that.
What I said was exactly right - both that they weren't the usual
songs and that it was going to be nearly impossible to find
them. They eventually did find the song clips, after being given
the song titles they occurred in and the concert links. But they
failed to find the posters, and so couldn't do anything with
that, and eventually gave up. Much more explicit clues to
Relisten and PosterDrop were added in later versions to help
teams have a chance, along with clips that have some lyrics to
make it easier both to identify which songs they are supposed to
be and to give a chance of finding where they occur.
We figured out the scientists clued were all women, and
simultaneously thought "Hey, that's cool to have a women in
science puzzle" and "Geez, the fact that women have often gotten
short shrift for their achievements in science makes these
people harder to look up." We did find some web sites devoted to
women in science that helped find some of our answers.
I helped this one out by using my knowledge of past hunts
(and my index) to find a bunch of shells for the hermit crab.
And then I was pretty much hands off after that.
Oh, Barney... I didn't see this one in testing but the joke
is hilarious.
I don't know how we found Barney Rafter, but he's a real
Australian who came to MIT all the way from his home country for
the Hunt. He played Charon in the opening skit and also was the
obviously Australian voice in the safety video. He turned out to
be a fun guy and also fully willing to make jokes about his
country being Down Under, both by representing it as the
Underworld in the opening, and by making this puzzle joking
about it being upside-down, or Australians viewing the world
with south up, or however you choose to interpret it.
Originally, this was going to be an app that let you enter
words like NYT Spelling Bee, with the space for the key word in
each puzzle highlighted and all the words listed alphabetically
to help you narrow in on it. There were two different issues.
First, the words we could use were pretty limited because of the
need to have them in real NYT Spelling Bee puzzles, which never
use S and can only have 7 different letters. At one point I
helped the first author eliminate a word on his list because it
had 8 different letters, and it was because of this we realized
the puzzle couldn't make the answer it had been assigned and we
had to ask for something else.
No other suitable answer showed up, so editors wanted to give
this puzzle the almost-thematic answer that was available and
find a different way to make the puzzle work which could get
there. (The original extraction was, after finding the winning
words in the interactive puzzles, to find NYT Spelling Bee
puzzles which could spell the runner-up words, and use the gold
letters from those puzzles, with given months of the NYT
Spelling Bee puzzles to provide the ordering.) And this is when
the second issue showed up; the author had become unavailable,
and we didn't have the tech resources to go in and write the
app.
We called in Jeremy Conner to rescue the puzzle, turning it into
a non-interactive version, and with a different extraction
mechanism which was true to the original concept but still quite
limited. He was forced to use weird bee from 1928 which didn't
follow modern rules and is misreported in places, as the
only way to get the letter B, but we couldn't come up with a
viable clue phrase that didn't involve the letter B.
One team complained that the clue phrase was ambiguous, but E.W.
Scripps' Wikipedia page reports only one half-brother. They must
not have made the connection that E.W. is the person who the
spelling bee you've been researching is named for.
This puzzle was improved significantly due to my test-solving:
Several tree images had details added to make them more
identifable, the flavor text got more cluing, and the tree collage
at the bottom was originally so badly pasted together that it was
impossible to determine what some of the trees were supposed to
be.
This puzzle's very close to the way it was when I test-solved it.
The flavor text was changed, and several captions were changed,
including one duplicated photo caption which was just plain
wrong...
I recognized several of the celebrities, but there were a lot
I didn't, so I fed them all into Google. When I had the entire
list, I noticed that they all seemed to be from a certain era.
Roughly speaking, it was my parents' generation (and recall that
this was my 25th Mystery Hunt, so I'm in the parents' generation
of today's MIT students).
But it was likely something more specific than that. I decided
to do the MJystery Hunter's equivalent to the old "I'm Feeling
Lucky" button on Google: I pasted the entire list of names into
a single Google search. And that gets you to The Muppet Show.
So we were able to get on to the part about names of muppets.
The version I tested used muppets from all sources, including Fraggle
Rock and some more obscure sources. Our difficulty in
finding them led it to being redone using only Muppet Show
and Seasame Street muppets.
I joined a testing session in progress for this one. My partner
had already figured out the basic idea, but I pointed him to the
logical conclusion.
He knew how to solve each room, and how the rooms connected
in a binary, exponential structure. So he knew the puzzle took
more steps to solve than you could even hope to have a computer
programmatically enter into the puzzle. I had experience some
puzzles like this, and said, "count the steps." And he's like,
"really?" But I helped him count the steps for each part and we
got to the answer.
This puzzle is largely the way it was when I testd it, but
several clues got minor tweaks. The two just-plain-wrong ones were
the "most populous city of some country" was wrongly called its
capital in the version I first saw, and the one that specified a
number of degrees was tweaked to be more accurate.
This puzzle was pretty similar when I tested it. Some numbers got
tweaked as a result of things we found. Even more numbers got
tweaked that we missed. It was that kind of puzzle.
Of course the fox and dog one got us started. I was reminded
of the "it
finally happened" video. Eventually we did the right
search to find one of the other pangrams and its English
translation. Some of the numbers were wrong because of math
errors, some due to mistaking letters with diacritical marks
that make them count as different letters in their languages'
Scrabble sets.
The big circle of arrows around the grid in this puzzle was added
as a direct result of my feedback. There were some other minor
tweaks as well. It wasn't too hard once we figured out what to do,
but figuring that out was too difficult without a couple fixes.
Yet another case where I was called in to get test-solving
teammates unstuck. They had already mostly filled the grid, and...
Identified VERLAN and L'ENVERS as answers, and found some of
the English words that could be verlanned. They needed help
finding more of these and throwing out candidates which worked
poorly.
We were trying to apply "difference" to the difference between
the sums of the clue halves for the original and verlanned
words, but we had enough wrongs in there that we didn't
immediately identify the shared half-clue in each of these
pairs. Weeding out the wrong ones (which usually didn't work
phonetically, didn't fit the supposed clue match, or both) and
finding new matches eventually made this apparent.
But in addition, we had one good but unintended match (the word
wasn't supposed to have a verlanned version), but it worked
phonetically and had a potential clue, though that clue didn't
share the numerical qualities of the others.
I also saw my teammates had had trouble with a numbering issue
which led them to reject a correct clue answer. One of the
numbers appeared twice in the grid, but since this wasn't a case
where the across and down entries started at the same number,
they had rejected the match earlier. Revisions to the puzzle
changed several answers to fix the red herring verlan and also
fix the numbering so no number appeared twice in the grid.
I didn't see this puzzle in testing. But I love it.
The song should be readily identifable to American solvers,
at least, even though it's almost 60 years old. Some may get it
from the title, and some from the posterboard photos, which make
an homage to the iconic music video that is well known in its
own right. It is, in fact, one of the very first music videos,
and perhaps the first that didn't just show the band performing
or people dancing to the music of the song. Some may recognize
it from Weird Al's spoof "Bob" in which the lyrics sung and
shown on the cards are palindromes.
But the puzzle uses an element of this classic video that has
lain in wait all these years from before the start of puzzle
hunts: The words on the cards include PIG PEN and WRITE BRAILLE,
two alternative alphabets puzzle writers love to use as ciphers.
Bob Dylan was yet again ahead of his time by including these
words in his lyrics!
I test-solved this puzzle consecutively with A More 6 U 28 U 496
U ... (which you'll see below in NYC) and so the test admin says
to me "More math!"
This was a pretty long solve, but it was never stuck; there were
sufficient guideposts all along the way. The puzzle was pretty
close to its finished form; there were a couple minor details to
fix (two integration variables were missing).
During a final check of this puzzle the morning before Hunt
started, I discovered that Wikipedia had two different versions
of...
The medal counts for Belgium in the 1920 Olympics on
different pages. And there were citations for both versions!
We edited Wikipedia to make it match the data already printed on
the physical artifacts we were distributing, but it's probably
wrong; what should be the most authoritative source has the
other set of counts. Now that hunt's over, I'm staying out of
it, even though I was the one who discovered it. We have a
Wikipedia admin who can try to straighten this out and handle
any resulting revert wars.
I thought this video was going to get redone so you didn't have
to listen to Hermes's horrible "singing" but it sounds the same as
the one I tested. Sorry.
Remember, you're in a Mercury car from the god Mercury? Of course
you can drive to Hawaii!
This puzzle was a huge challenge in testing, but even worse than
it was for you because we didn't have that flavor text. Even
knowing what you are supposed to do, it's a huge challenge. If you
missed this one during hunt, and you can find a group of
multi-lingual cryptic lovers (or are good at interpreting Google
Translate results), give it a try, but allocate a good amoutn of
time for it.
The problems with this puzzle started before we even got going on
our test-solve. The playlist was originally shared as a real
Spotify playlist. I clicked the link, and Spotify gave me a 404.
They asked me if I was logged in, and I said no, I don't usually
use Spotify. But Spotify thought I had an account (no doubt I did
something like this before) and I reset the password and logged
in. Clicked the link, 404.
I confirmed I could listen to songs on their suggested playlists,
and I could make my own playlist (I searched for and put one of
the songs from the puzzle into it, which they had sent me as
screenshots by this point). But I could never see the list the
author provided, even though some other solvers on our team could.
We never figured it out, but I assume they found other people who
couldn't view the playlist either, leading you to get the image in
the puzzle instead of a real playlist.
For actually solving the puzzle, I floundered a bit at first.
The puzzle was called Show Tunes at that time, and since the
first three clues mention reality shows and I assumed the masked
competition one was also one, I looked for these songs performed
on reality singing shows. This gave me plenty of time to fully
read through the list of songs, and I thought the distribution
of songs seemed about right - some new ones, a lot of greatest
hits from across generations.
At some point, I got down to "I Dreamed a Dream" and I didn't
have to google that one. I remembered this song was sung in one
of the most famous performances on any reality singing show
ever, when Susan Boyle sang it on Britain's Got Talent,
and sang so well the performance got repeated around the world.
Convinced that was supposed to be the right performance of that
song, I checked her against all the clues. The only one that fit
was that she wrote an autobiography.
The next match I found that seemed really solid was a song
performed by the winner of one season on The Masked Singer.
It was easy then to look up all the songs Masked Singer winners
performed on the show, and there were a total of four songs
performed by three different winners. I expected this: We had 55
songs and only 17 clues. I wasn't sure how it was supposed to
work after that step, but I was making slow progress. I assumed
flavor text (which was later changed) saying the playlist was
"pretty much perfect" meant the songs got the highest ratings
from judges on the various shows, and that was to help me in
cases where I found different performances of the same song.
My absent solving teammate showed up overnight, and figured out
the puzzle was supposed to be about songs sung on Glee.
Every song in the puzzle had been sung on Glee. This was
disappointing compared with what I already found, but it was
reassuring in that I had a much tighter set of data to work
with. There was a bunch of data collection and checking and
rechecking here. One song in the puzzle was replaced to fix an
error. Also, the clue about dating, which originally just said
"I dated somebody else here," was changed after I sent feedback
with a list of Glee actors who had dated other Glee actors;
there were a lot more than the author realized.
We eventually solved it the way it was meant to be solved. There
was one other thing... that answer. Some of the solvers during
the actual hunt also failed to recognize that was a real thing.
We were missing the last letter, because of not understanding
how that clue worked, and I googled something with a different
letter there and Google corrected it to the right answer.
This puzzle was mostly the same as you saw in Hunt when we tested
it. The interactive bit was a Python script running through some
Python-web gateway which might have been the same as what's in the
puzzle now too, except we had to type the names of the cards
rather than use buttons to pick them.
At some point, my teammate found a way to crash the Python
script. The script was fixed and we continued.
This was the biggest BE NOISY moment I encountered in testing
this year's Mystery Hunt. (For the uninitiated, this
happened during the endgame of the 2002 MIT Mystery Hunt.)
I joined two solvers who had already figured out that the
grid had references to Parks and Recreation episodes,
and the images on the tiles referred to SubPar Parks, and the
text on the tiles was mashed up from both sources. They were
stuck on how to use the symbols on the edges of the grid hexes.
They thought it was going to be something related to Catan
because of the board game reference in the flavor text, and the
19 hexagonal tiles. In the version we were testing, the tiles on
the map at the end were arranged in a hexagon, exactly like
Catan. But they had never played Catan and I was added to the
session to give them more board game knowledge... when that
wasn't what the session needed at all.
But anyway, they thought that the 4 different markers might
indicate possessions of four different players. I analyzed it
from a Catan perspective. One of the markers pointed to a vertex
and the rest to edges, so i took that as a settlement and some
road segments. But I realized the Catan rules only allow each
player to build up to a maximum of 15 road segments, and since
they can only build out from the settlements and other roads,
and can only add more settlements on roads they have built, each
player can have at most 2 disjoint road networks. One player had
8 of the marked segments, and they were too far apart to join
into only two networks using 7 more segments.
Then I came up with the BE NOISY. The marked segments were well
distributed, only in two cases with pairs adjoining at the same
vertex. I came up with a set of rules for a logic puzzle: The
marked vertex is an endpoint for the path. Each marked edge is
on the path. The path never touches itself, so at most two of
the three edges at a vertex can be used. The markers were all in
different hexes, so the order those edges and one vertex are
used is the order of extraction, and the different symbols
indicate extraction indices somehow. Here is the map
(spreadsheeted) and here is the solution
(black lines form the path) to that puzzle, which I proved
unique.
The only problem is that none of that was intended, and after
our extraction attempts for that path all failed, we got an
update which gave the "blaze a trail" in the flavor text
(original was "retrace a path"). After learning what those
little symbols actually were, and figuring out how to use them
to make the intended path through the hexes, rather than along
the roads, we solved the puzzle.
Since there was no actual board game involved in the puzzle, I
was upset at the massive red herring we'd been fed. My feedback
included "I guarantee about 90% of Mystery Hunters who look at
the current form of this puzzle are going to think of Catan, and
the other 10% simply don't know Catan and their teammates will
get them headed that way. The first two solvers ... hadn't
actually played the game and they still recognized it as Catan.
And in fact, this puzzle has zero to do with Catan."
I advised them to dump the board game references entirely,
advice which was not taken, but at least they rearranged the
hexes so they weren't in the pattern of a Catan board. The
actual board game reference was supposed to be Cones of
Dunshire, a fictional board game that appears in two of the Parks
& Recreation episodes not otherwise used in the
puzzle. Apart from having a map made of hexes, it has no
resemblance to the puzzle and no influence on the solution to
the puzzle. But it looks like we got no Catan-related answer
submissions and only two teams who mentioned it in their hint
requests, so I guess the changes worked, mostly.
I found this puzzle fun but long. I broke into it in a manner
similar to but not exactly like what is described in the solution,
and once I did worked mostly from top to bottom as described in
the solution. And I had a long nitpick session with the author
afterward to iron out lots of little details that were not quite
right in the version I solved.
One of the teams registered for this hunt was named
ℙoNDeterministic, using a blackboard-bold P as is used in this
puzzle to represent a set solvers have to figure out. I wonder
what theirs represents.
This was one of the most impressive puzzles from our hunt. Elaine
Lowinger built a completely custom pinball machine for solvers to
play after they solved the first part of the puzzle. Until now, I
had assumed I was the second-biggest pinball nut on our team
(second to Bowen Kerins, multiple-time World Pinball Champion, so
second by a big margin, but still second). This makes me think I'm
third.
Elaine lives in California, so she had to build the machine and
get it working, then take it apart to ship it to MIT and put it
back together in our headquarters just before hunt.
And Bowen discussion-edited the puzzle, which had to be
incredibly cool for Elaine. Imagine you decide you want to write a
chess puzzle and your team says Garry Kasparov is going to be your
advisor on the project. It was like that.
This was a puzzle I discussion-edited, and it was one where my
discussion-editing included writing a program to prove uniqueness
of a part of the solution.
My other guidance for the author was that solvers are going
to find the parity errors confusing and difficult, and they
should try to keep the other aspects from getting too
complicated so solvers don't just abandon the puzzle in
frustration or confusion. This led to simplifications like
getting the original Rubik's Cube colors which matched in the
standard Rubik's Cube way instead of also having solvers have to
figure out the arrangement of colors on the cube as part of the
cube assembly (even though my program had proved uniqueness of
that).
With this puzzle, I mostly helped guide the early stages of the
work, pointing out that if the Magic-chess interaction was too
loosey-goosey then it would be too hard for solvers to confirm how
the puzzles were supposed to work. I also poked holes in some
proposed puzzles by finding alternate answers. Later on, editors,
factcheckers, and test-solvers were filling that role and I was
mostly hands off.
I encountered this puzzle as a test-solver, and it was a
nightmare.
The version I test-solved was over twice as long, and there
were a lot of issues. The clue for Astana was there about its
world record (for most renamings of a capital city in modern
times) which was a good break-in that the puzzle was about city
renamings, but some of the renamings were dubious. Some could
only be found on one Wikipedia page, while other pages solvers
might be looking at for solving the puzzle glossed over those
changes. Another issue was that some of the clues referred to
the events that led to the renaming, but not the actual name,
while some of them only referred to the name, not when it was
renamed, sometimes making the dates seem out of order.
After I finally got through it the author let me know he was
carrying on the work of another author who hadn't had enough
time to keep writing the puzzle, and a lot of the issues were
due to trying to preserve the original author's work. I told him
that if he was in charge of the puzzle now that he needed to do
what it took to make a good puzzle. Much like the "Kill
your darlings" advice, he should verify all the data the
puzzle was based on and cut out dubious and just plain wrong
renamings, and double-check dates (in a number of places I'd
pointed out). Also, his clue phrase was awful; that was another
holdover from the original author. The editors helped with that;
by assigning him a different answer, he was forced to choose a
new set of cities to extract from and was able to implement my
suggestions.
A fun one. I especially enjoyed the teleporting express trains
(just look at the timetables).
I solved the subway routing puzzle by hand. After thinking
about how I'd write a program to solve it, by tracking the
quickest arrival time at each station and when a better one was
found, updating all the other stations you could now reach
sooner from there, I just did it by hand with that algorithm.
My teammate found the Clefs d'Or link and then we were able to
solve the puzzle on the message board.
When I test-solved this puzzle, being added as reinforcement
to help the original solvers clean up their data, after we got
most of it cleaned up, one of my teammates found the final clue
phrase ROO/SEVE/LT/ORB/EAR among unused images, and we
short-circuited the puzzle. We were supposed to get a first
message from the unmatched images near the start of the puzzle,
which was less clear for us because we had not connected some of
those to their sets. Since the part we skipped was solving
number links, it was just as well they left the puzzle like this
where it could be short-circuited.
One of those should-not-be-missed puzzles. Many will recognize
the obvious homage to the late Jack Lance's OCTOGRAM game. Jack
died suddenly during the construction of our hunt, and while I
wasn't as close to him as some Mystery Hunters were, I had
experienced some of his creations and he was one of the few
puzzlers I could say always kept me in awe. We will never see what
Jack Lance might have done as a Mystery Hunt author, which I am
sure would have been astounding. Right after his death, my
teammates were saying they wanted to come up with a fitting
tribute to him, and I think Jim Hays made a fine one in this
puzzle.
I noticed (as
did Alex Irpan) a flyer (completely coincidental) posted
on one of MIT's numerous bulletin boards about the pensonality
styles enneagram, which was used in this puzzle.
This was another puzzle I discussion-edited, and another where I
wrote a program to help confirm uniqueness of part of the puzzle's
logic.
In this case, I confirmed that even when you repeated pieces
in different grids, considering only the shapes and immediate
contradictions due to placements of the same number, only the
sudoku grids that were intended could be made.
An earlier version of the puzzle which had more, smaller pieces
and didn't yet have the every-digit-at-least-once-per-piece rule
had trouble with extra grids that were hard to rule out until
you tried fully solving the sudokus.
I joined an existing test-solve session of this physical puzzle.
My teammates had the pieces and had assembled all but two of the
images when I started. I should point out that instead of getting
20 strips per painting for a total of 160 strips, they had the
strips cut into squares, so they had over 2000 little squares to
assemble... and that was too much, so it was toned down to the
strips you got. Jigsaw
Puzzle from the 2013 hunt gave you even more pieces than
this... and it was too much then, too.
The photo sides of these jigsaws showed famous oil
paintings. The back sides had a lot of letters. There were small
letters that didn't spell anything, in any direction, and larger
letters that spelled words reading horizontally - the same word
over and over horizontally, not necessarily starting with the
first letter in a row, and they had written down all these
words, and did so for the last two as they got them assembled.
They had two things: Given the oil theme, they noticed a number
of the words (but only about 1/8 of them) were types of oil or
words that made a phrase "___ oil." And they noticed one
painting had an image of a dragon added to it that was not in
the original.
I looked at the pattern of where the oil-related words were, and
while they tended to be spaced apart, the spacing was irregular
and different for each one. It wasn't obvious what it meant.
So then I got my team's attention focused on the dragon. Clearly
that was not there in the original painting. So we should look
for things added to the other paintings. This was indeed what we
were supposed to do, and we started finding the hidden animals
in the other paintings. I spent a while trying to find the exact
Monet Water Lilies painting used in the puzzle; he
painted over 250 paintings in that series, including 18
featuring the bridge seen in the one in the puzzle, and it still
took some effort to identify the animal. I have since seen this
picture life-size from the strips we gave solvers, and the wolf
now pretty obviously isn't part of the original. I think it was
redone to make the animal clearer, and it came out so obvious
that it would likely be identified at the start along with the
dragon. Comparison of the version
solvers got (left) with what I had to work from during
testing.
But at the end of this stage we had a list of eight animals
including a dragon.
Our work paid off, because with the list of animals and "oil" my
teammates found... [continued next spoiler box]
The other important thing in this puzzle is that we got Franced.
Those of you with good memories may remember that the last time my
team ran the Mystery Hunt, we discovered days before Hunt that France changed
its regions on that January 1st, breaking a puzzle we had
written that involved them, and forcing us to find another
suitable country to put in their place. The details of who Franced
us are inside the spoiler block, but since this puzzle was already
professionally printed on light card stock, there was no fixing
it. We just told solvers to use the 2023 data. And now the part of
the solution that got Franced:
... the list of oil patterns used by the Professional
Bowlers' Association. And we learned a lot about bowling that us
non-bowlers/occasional casual bowlers never knew. Bowling lanes
are oiled to ensure the balls glide down them. There would be
too much friction for the ball to get all the way down there,
otherwise. But they aren't oiled uniformly. Different
parts of the lane get different amounts of oil. This affect how
the ball breaks when you apply spin to it when bowling the ball,
and it is how professional bowlers bowl those curve balls that
hit just behind the head pin on either side in order to make
strikes. And all lanes aren't oiled the same way. At
your local bowling alley they probably use a "house" oiling
pattern which is friendly to bowlers who have a basic
understanding of how oiling works, and you can ask to see a
chart of this pattern.
In PBA events, they use one of a number of named patterns, which
are named for the animals in our puzzle, and which are
considerably meaner. Consider how a newspaper crossword compares
to a crossword you get during Mystery Hunt. These patterns
change the amount of oil and distribution of it across the width
of the lane and certain distances down the lane, with these
breakpoints measured in feet. The lane is 60 feet long and the
grid on the letter side of our jigsaws is 60 squares tall, with
the large letters spelling words placed on these foot-marks.
Sure enough, there are oil-related words on most of the
breakpoints in the patterns named for the animal added to the
painting on the other side of each jigsaw. One on each jigsaw
isn't, and that word is the extraction to reach the next step.
Or, rather, they matched when we test-solved it. For 2024 the
PBA changed the patterns so they no longer matched our puzzle.
I saw quickly that it was about The Count, but I didn't get
that it was from the specific album. Searching for the largest
number 2478693 and "the count" gave as the first hit a Sesame
Street fanfic script for an appearance of Britney and
Jamie Spears as guest stars; not a real script, but based on the
lyrics of "Beep." The actual song lyrics don't show up, but I
found "2,478,693 beeps" in another very random list packed with
Sesame Street references, so I figured it must be from
the show. Searching for a sequence of numbers near the two
15000s gave a page which quoted "The Count's Weather Report." So
I ended up listening to a bunch of Sesame Street skits
with the Count and getting one or two more of the right ones. My
testing teammate stumbled upon a playlist from the album and so
suddenly we went from a couple to all of them as quickly as we
could match them up.
This was a long, difficult test-solve because the puzzle was
explained poorly.
The first part to assign values to letters based on the sums
given when you solve each puzzle wasn't hard. But the original
puzzle text wasn't clear about wanting to ignore the actual
motion of the stars, and not knowing how to use the numbers from
the strings, we thought we were supposed to figure out where the
stars would be in 10 million years, compare the coordinates to
the numbers from the strings, and derive something that way.
(The fact that we didn't have the interactive component yet hurt
as well. We were just told it would exist.)
And using the real data didn't work at all. Compared to the
stars we were really supposed to be looking at, the real stars
in the constellation are much closer to us, and have greater
apparent motion in the sky (for comparable actual motion) as a
result. The apparent motion of the real stars is so great that
it doesn't make sense to assume that the annual movement,
measured in terms of changes in two angles from spherical
coordinates, would remain constant for so long. Some of them
would have revolved completely around the sky in that time!
We finally figured out what it was supposed to be after we got
the flavor text that was used in the end, that Gaea didn't like
the way they're changing... what if they changed (some other
way)? And we still didn't get it immediately, but we figured it
out. What slowed us on this part was that even though we had
retrieved some of the real star data from the GAEA project, we
had recorded every identifier of those records but the one the
puzzle was trying to match by. The numbers they gave us are
searchable, but we didn't realize the database had identifiers
in that format.
It was immediately obvious upon opening this puzzle that it was
based on...
Who Wants to Be a Millionaire? It wasn't until my
teammate did the right search that we figured out it was also
based on the song "If I Had $1,000,000" by Barenaked Ladies.
A little research showed that since I stopped watching the game
show, they had changed the dollar values a few more times (I
remembered the first change) and most recently they were back to
the original values. And only those original values included
$32,000, so those had to be the right numbers. But how to use
them?
We had an acrostic from the questions when we put them into the
order of the song: USE RIGHT ANSWERS. This made me think I was
supposed to use the A, B, C, or D of the correct answers to
extract one of the first four letters. Seeing this wasn't
working, they asked us to consider if the acrostic had instead
said KEEP PRIZE ON AYES. That still didn't help, so they tried
DIGITS AS INDICES. Finally it made sense. Without that clue, it
was "guess what the constructor is thinking" to figure out that
the $32,000 level meant to extract the 3rd and 2nd letters of
the answer.
We also had a couple nitpicks in the questions. The empire
spanning one included Dutch and Portuguese in the version we
saw, and we provided evidence to indicate both of those empires
had spanned six continents. The Hannibal question implied
Hannibal rode on an elephant across the alps. While he did bring
elephants with him, they were for use as war machines.
Historical evidence suggests he rode a horse during the
crossing. And they added more BNL references in the questions
while they were at it.
You could probably guess this was one of the puzzles added late
in construction...
Since it was based on a video game expansion that was
released in full on November 21st before hunt and in public beta
just two weeks before that. So test-solving it was complicated
by not having good resources for the the content that was just
released, and actually got better a week or so later when we
found better-summarized guides.
I didn't see this puzzle in testing, or at all until hint
requests came in for it late in the hunt. And those hint requests
tended to stick around because even the hinters had trouble
understanding the puzzle. The solution page is hard to follow. But
I took a good long look at this one after hunt. I think it all
works, but it's hard to follow because it doesn't work how you'd
expect. It's superficially a programming puzzle, but it doesn't
really work like one. Here is what you were supposed to figure out
(inside a huge spoiler block).
There are three code snippets labeled test1, test2, and
test3. You're supposed to figure out from red-highlighted
keywords in each test program that they are written in Python,
Ruby, and Go, respectively. The file extensions missing from the
filenames are py,rb, and go. The files
are supposed to fit into the logic gate diagram at the spaces
marked 1, 2, and 3. You see that the output of 3 is already
labeled with the letter O. You're supposed to figure out from
this that you assign the two letters of each extension to the
input and output of the numbered gates, so for example the
maroon trace heading into gate 3 is assigned G. That said,
Python's io module doesn't have io.bind() and much of the rest
of this code can best be treated as pseudocode. But you can
interpret each piece in conjunction with the colored logic gate
diagram.
The function in each code snippet begins with a comment
referring to documentation from a particular date in the 20th
century, ending with "Comments needed here!" which literally is
a "request for comments." In addition, each of these comment
sections has the acronym RFC. Internet standards were published
as RFCs, requests for comments, in that era. The first section
of each test function binds to a well known network port: 21 is
FTP, 25 is SMTP, and 80 is HTTP. The dates help you disambiguate
which RFC for each of these services you're looking for: RFC 114
for FTP from April 1971, RFC 788 for SMTP from November 1981,
and RFC 1945 for HTTP from May 1996.
Each test function then reads (io.recv) a value from a
three-byte hex string. Hopefully the various colors and the
author name roy g. biv made you curious enough to extract the
exact RGB color codes from the logic diagram. If you did then
you'd have found that the hex code arguments to recv are the
colors of the traces connected to the three network ports in the
diagram, the green, brown, and pink traces (in that order).
Finally, each test function asserts that a particular value was
received. These numbers are close to the corresponding RFCs, but
each is too large by a two-digit difference. The comment refers
to codes on page 38; this is a reference to "code page 38." The
DOS/Windows code pages have larger numbers than this, but the
code page system started with IBM mainframes, in which code page
38 is US ASCII. If we interpret the difference as a character in
ASCII, this gives us A for the green trace, I for the brown one,
and T for the pink one.
The next part of each test function sends data and then reads
more data (test2 and test3 each do this twice). As with recv,
the groups of three hex bytes should be interpreted as colors,
and each of these sections corresponds to the traces entering
and exiting one of the AND or OR gates in the diagram (the two
gates at the left are coded first, then the three on the right,
both sets top to bottom). The two-byte hex strings are the data
sent. This time, code page 37 in the same nomenclature is
EBCDIC, the more common character encoding used on those
mainframes, and this is confirmed by the first three bytes of
the asserted value.
When you interpret each set of four bytes of sent data as
EBCDIC, you get four capital letters spelling a word, which is
meant to be a clue. For example, the section in test1 sends
0xD3D6, 0xE5C5 and this spells LOVE. The answer is formed from
the letters of the two traces feeding into the gate (starting
with the top one), then the name of the gate, and finally the
letter corresponding to the output. Each asserted value ends
with some null bytes; the total length of the asserted string,
in bytes, tells us the length of the answer we are supposed to
get. By working back and forth from the traces we already know,
we can figure out the words as ADORE, ADORN, BRANDY, GRANDE, and
IVORY. Now we have the letters for all traces.
The last code snippet has its file extension, kt, identifying it
as a Kotlin program, though its comments tell us it's
incomplete. It calls a function exorbitantHash, which is
supposed to clue that it's using XOR. The first three bytes sent
as arguments to this function make color codes again, the brown
and red colors from the diagram, the ones assigned I and O in
the original puzzle diagram. The fourth argument in each call is
an ASCII letter (I or O). And the last bit indexes into the
discriminator array; this is a 0-based index, so [8] and [14]
grab the 9th and 15th elements of the 26 byte array,
corresponding to the positionas of I and O in the alphabet.
Putting all this together, we XOR the three bytes of the color,
the ASCII value for each letter assigned to one of the colored
traces, and the corresponding byte from the array. The resulting
byte is mostly 0s in binary, with zero to three 1s. There are
only one or two 1s in each bit position across all these
letters. cluing that this indicates which letters can go into
each position of our answer. For instance, the operation on I
XORs 0x9a, 0x63, 0x24, 0x49, and finally 0x1c from the array to
give binary 10001000, indicating I can go in the first or fifth
position. Only two eight-letter words fit the indicated pattern,
forming the answer: INTERNET PROVIDER.
The "modernized" Romeo & Juliet prologue was funny right from
the start, with dead-ass props, people dissing people's Pops, and
being advised to turn your cell phones off and watch the show.
But the Shakespeare just kept going deeper and deeper. There
are a bunch of other Shakespeare characters from other players
listed to the right of the prologue. We figured out they are the
speakers of the "interruptions" in the prologue, which was
originally a sonnet but is now more than twice as long, and
matched them up. Where we didn't get them just from the
paraphrased text, we looked up all lines from each character.
But some of the characters were ambiguous. "Ham." actually did
turn out to be Hamlet, but "Jul." wasn't Julius Caesar, but
Julia from Two Gentlemen of Verona.
You should appreciate the fact that we test-solve metas with one
or more answers missing.
In the first version of the puzzle we got, Theseus and the
Minotaur were swapped, and this led to us trying to build the
path in the wrong direction, which still let us think about
possible paths with the ship sections we had, but the extraction
was backward and hence hopeless. In addition, we didn't have the
beef grades, and were trying orderings based on pulling good
letters out of the answers, so we were using the wrong indices
to extract from the answers in the wrong order. An update
swapped Theseus and the Minotaur on the maze, and after we
figured out the beef grades, we still had trouble. With two
answers and hence one pair of maze pieces being withheld, we
backsolved the missing answer corresponding to Canner to get
those missing pieces. (There were other testers, so don't think
this was the only way to solve it.)
