Ella Musk's Field Trip
Kiera O'Flynn

This puzzle involves five subpuzzles, each extracting three letters for the final answer. For each subpuzzle, you will need to identify the names of three characters, transliterate text written in the region's script, and solve a (logic, music, etc.) puzzle.




Text: YAJILIN spells out "YAJILIN" in Inazuman script.

Puzzle: From the script, we know that we're supposed to be solving a Yajilin puzzle; on top of that, according to the instructions below the logic grid, we want the solution loop to travel through 1->2->3->1->... when read clockwise. Clicking on the logic grid, however, reveals that there's something unusual about this grid. Instead of just shading/unshading the clicked cell, the cells seem to be on a 1->2->3->(shaded) cycle, where toggling a cell makes it go backwards one in the cycle and makes its immediate neighbors go forwards one in the cycle.

With the mechanics in mind, then, we start solving. (This is just one of many solve paths!)

Just like in a normal Yajilin, we know that the two shaded cells can only go in these locations:

Since there's only one way to shade these cells, we know that the cells must be toggled and their neighbors must be untoggled.

The last cell on the the third row must be a 2 or a 3; since the last cell on the fifth row cannot be a 1, this cell must be a 3, and the fifth-row cell must remain a 2. If the fifth-row cell is still a 2, then its neighbors must be untoggled.

From there, we can deduce the states necessary for the 1->2->3 path to be preserved:

If the path currently at R3C5 was to travel downwards, the cell would need to be a 1. Since that's not possible, the cell needs to be shaded.

Then, if the path was to travel leftwards (or if R1C5 isn't shaded, or if R2C5 is a 3 instead of a 1), we quickly run into a contradiction, so we can deduce the following path:

Since R3C4 can't be shaded, it must have a path through it.

And since it can't be a 2->3->1 path, it must be a 1->2->3 path.

Similarly, R2C3 can't be shaded, and it can't connect to R3C3 since it can't be a 1, so it has to loop around like this.

Filling out the loop:

R3C2 now also can't be shaded and can't connect to R3C3. The reasoning from here on out is pretty straightforward.

Finally, with our solution in hand, we fill out the across/down character names and extract the shaded squares in each name in order to get HOR.

Sumeru (Rainforest)



Text: There's nothing obvious to transliterate yet...

Puzzle: Finding the text in the image!

Starting with the leftmost column, we identify Nahida's locations (pointed at by the arrows) in the four pictures in the main image, finding the letters FOUR, or FOUR in Sumeru script.

Next, looking at the middle column, we see that we need to identify some Tri-Lakshana creatures. In the image, we find a starshroom, floating seed thingy, rocklifting branch, and the same bouncy mushroom from Nahida's image. Clicking on them reveals some writing:

Ordering them by the number of lines, we get the letters FIVE, or FIVE.

Finally, in the rightmost column, we see that we need to find a Rukkhashava Mushroom, Zaytun Peach, and Harra Fruit. And find them we do:

These contain the letters TWO, or TWO.

Finally, indexing these numbers into each character's name, we get IZO.

Sumeru (Desert)


  2. DEHYA
  3. CYNO

Text: We have the individual letters N (N in Deshret script), O (O), and M (M) on the left. We also have the formula 2^{|N|+26} MOD 101 on the right, which we transliterate to 2^{|N|+26} MOD 101.

Puzzle: Inserting the characters' names into the formula, we get that 2^{7+26} = 35 mod 101, 2^{5+26} = 34 mod 101, and 2^{4+26} = 17 mod 101. Placing these numbers back on the obelisks, we get N35, O34, and M17.

We then look these up in Gardiner's sign list, getting the hieroglyphs that phonetically correspond to /n/, /z,s/, and /i/. Translitering these sounds back into the English alphabet, we get NZE (where the latter two letters are pronounced as in ZERO).




Text: We can transliterate each line of text above the blank squares from Fontaine script:

Puzzle: Each line references a song played in Fontaine and provides a timestamp. The goal is to recreate the song at that timestamp by dragging notes from the bank at the bottom to the blanks (purple onto yellow, and yellow onto purple, in accordance with the Arkhe mechanic). In the end, we get these matchings:

These results look very similar to the names of the characters. To be precise, these results replace each instance of E with R, I with O, and L with D. The letters being replaced are what the characters "have in common"; they're the only letters shared by all three of them. Ordering the source letters alphabetically as suggested below, we extract ROD from the replacement letters.



  2. KAEYA

Text: We can transliterate each clue from the Abyss script:

Puzzle: Each line, itself written in Latin, clues a Latin answer: ALBATUS, CAVEA, and DANDUM, respectively. The answer shares letters with the corresponding name (e.g. ALBATUS/ALBEDO) precisely where "O"'s occur above the character's head, and differ at the "X"'s (with the exception of CAVEA/KAEYA, since the K sound was represented with Cs in classical Latin). Extract the first letter of difference/first "X" letter (again with the exception of CAVEA, since the V is pronounced as a W) to get AWN.

And we're done! :)

Author's Notes

Shoutout to all the solvers (especially people who solved Fontaine, my favorite subpuzzle :>) for making this puzzle feel worth it. It made me kind of sad to see all the people who said "I don't feel like doing this puzzle because I don't play Genshin" when the puzzle was written specifically to be accessible for people who've never played before, but what's done is done, I suppose.

And thanks to the playtesters and my ethics course for making this all possible :)

These notes are going to read kind of like a blog post about the puzzle writing process, which I don't know if author's notes are meant for. However, I don't have a blog or anything, so you're stuck with this here.


WARNING: The puzzle might be spoiler-free, but the notes ahead contain minor spoilers for Archon Quests through Chapter IV Act IV, moderate spoilers for the major World Quests in Sumeru and Fontaine through 4.1, and the full puzzle mechanics for various regional puzzles. If you don't want to be spoiled on them, stop reading now.


I had the starting idea for this a while ago, maybe around mid spring. I wanted to make some sort of Genshin puzzle, but almost everything in the game is a spoiler (in a very loose sense) for something in the game. The one thing I could think of that wasn't a spoiler was the Aranara language (though a spoiler for Aranyaka, I suppose...) and the scripts scattered around Teyvat, because players aren't expected to actually ever figure out the scripts, and when they do, the process of figuring out the alphabet is really, really, really cool and impressive. To this day, we haven't even figured out the Aranara script (though I'm not convinced it's interpretable in the first place).

Anyway, that was the idea; however, I didn't have any ideas on how to actually use the Aranara/scripts, at least until Summit created a meta feeder answer specifically for a Genshin puzzle. I was so happy when she gave up on her puzzle idea and let me write the puzzle instead :>

In addition to having "write a Genshin scripts puzzle" as a goal, I also wanted to write a puzzle that included all of my favorite characters (Albedo, Elynas, Nahida, Kazuha, Arlecchino, Dainsleif, and Heizou, in that order), or at least as many of them as I could. Thus, the idea of associating regions' scripts with their characters was born, which also fit perfectly with the answer, since the answer is a "region" home to a "Genshin character". With the prominent Teyvat linguist Ella Musk being from Mondstadt, the puzzle name and flavortext shortly followed. (Originally, the flavortext was just the Hilichurlian without the translation! However, I really didn't want people to be stuck translating Hilichurlian when that's not the focus of the puzzle.)

Me trying to doodle my puzzle into existence

I then set out to write my subpuzzles, where each subpuzzle 1) corresponded to a region that 2) used a script, 3) used characters from that region in a spoiler-free way, 4) involved the language associated with the script, and 5) had a distinct mechanic based on actual puzzles in that region.

Title page

The map was fun to draw, but the grow-on-hover links were not very fun to program.


This subpuzzle was the first I made, and the idea came to mind almost immediately. Despite having a logic assignment due that night, I spent hours just writing out the script and admiring my work. It really sparked joy. Not only could I put Albedo in a puzzle, but I could also put my six years studying Latin to good use, and looking at the completed puzzle just felt right.

The slow process of lettering. (The "22" was for VV => W, but it was more confusing than anything)

Finding the clue answers was somewhat hard, and I wish I didn't have to do the awkward thing with Kaeya and the C/K V/W pairs, but the solution letters were fixed (since I wanted to "visit" the regions in canonical order), so I coped. I'm very glad I found cavea in the end, since my alternative before that was wainscotum with Dainsleif (which isn't classical Latin) and caenum for Kaeya (which means slime/filth, which isn't very flattering for Kaeya). I was very happy in the end with the both alliterative and character-thematic clues.

Thankfully, choosing the characters was easy; there are only really three playable characters associated with Khaenri'ah. The only awkward thing is that we don't know if the so-called Abyss script is actually used in Khaenri'ah, since Khaenri'ah won't be released for another few years. I was tempted to just have this region be "the Abyss" instead of involving Khaenri'ah, but the only characters from the Abyss are Albedo and his siblings, and at that point, it would make more sense to include Childe than Kaeya (and neither Childe nor Tartaglia are very Latin-friendly names, and looking up his wiki page right now would probably contain major story spoilers for people who actively play but haven't caught up with the most recent Archon Quest).

This puzzle is inspired by the most common (?) usage of the Abyss script:

A sealed chest.


This one also came really easily! Since I wanted to use romanized Japanese somewhere but don't actually know Japanese, I figured the easiest thing to do was to just make a logic puzzle based on a genre with a Japanese name. Since Inazuma is practically overflowing with puzzle mechanisms, it was hard to choose which to use, but I settled for a combination of these two:

The goal is to make a path which follows a M-K-T cycle, as clued in the below map.

Map! (Here's where the cycle clue comes from.)

You can find these things all over Inazuma. They all operate on a 1-2-3 cycle; hitting one will increase it by one, but it will also affect its neighbors in some way. The goal of the puzzle is to figure out how its neighbors are changing, and once you've figured it out, to set all of them to the same number.

They even made an actual logic puzzle out of them! It's a magic squares grid. It only made sense, then, for this to inspire my actual logic puzzle.

I started construction by choosing the characters. I really wanted Kazuha and Heizou in there, and luckily enough, they had just the right letters needed. I made a 6x6 grid with the two of them at the top, adding Kujou Sara right below them (since the only other Inazuma characters with an R in their first name were Gorou and Kirara, I missed Kirara at the time, and Sara worked nicely as a down instead of needing three acrosses). I mistakenly thought that Heizou's name ended with an O, though (which I didn't see until the first playtest... luckily I could just move the names around, but things would've been so sad if I had to reconstruct), which constrained the top right corner to be a clue with two shaded squares next to it.

Initially, I was going to stay true to the Inazuma mechanic and have the cells make full cycles instead of just toggling. However, that was way too hard to keep track of and didn't lend itself to interesting logical deductions, so I stuck with toggling instead.

Not gonna lie, I don't remember much of the construction process from there, but it was pretty hard; at some point, I wasn't even convinced there was a possible layout of starting tiles that could lead to a unique solution where no squares are shaded initially. I first tried setting all the initial tiles to 1, but that was unsolvable, so I gradually adjusted the tiles until there was an actual solution. At that point, I hadn't implemented the nice HTML logic stuff, so construction involved drawing over the grid with a web paint extension and refreshing & starting over every time the initial tiles needed to be updated.

In the end, though, at least, I was really satisfied with the result. This puzzle felt very Inazuma.


This is where my puzzle starts being sponsored by my ethics course.

My first idea was to do some sort of interactive maze sort of thing, since the only puzzles in the desert I can think of are 1) everything in the ruins and 2) the weathered obelisks I ended up using, and as someone who 100%'d the Hypostyle Desert without using the interactive map (and before the underground map released), getting lost in a pyramid and doing Deshret's ruin puzzles while literally having no clue where you are feels very, very desert. The only problem with that was the difficulty of construction.

I was in an ethics lecture, then, doodling on the handout they gave us, when the idea just came to me. It wouldn't be as satisfying of a puzzle, but it would still feel pretty desert, and at least it would work (and hopefully be pretty easy, as desert puzzles are). The idea was to extract to 𓉖𓎁𓈗, which resembles (?) NZE.

A first draft

Collecting images to use

The actual testable draft I procrastined putting together for three weeks

The resemblance was not enough to actually get the letters, so I ended up switching to a phonetic rather than visual extraction.

Hi, my name is Nze

The original formula wasn't too hard to come up with, since the characters were fixed (there are only 3 characters from the desert) and the Gardiner numbers were all convenient multiples of five that had a nice correspondence with the characters' name lengths. However, 17, 34, and 35 was much harder to find a formula for. I ended up consulting OEIS, and I'm really glad I found the formula I did, because the alternative sequences were all things like "When dealing cards into 3 piles (Left, Center, Right), the number of cards in the n-th card's pile, if dealing in a pattern L, C, R, C, L, C, R, C, L, C, ... [as any thoughtful six-year-old will try to do when sharing a pile of candy among 3 people]" and "Lexicographically earliest infinite sequence such that a(i) = a(j) => A000593(i) = A000593(j) and A336158(i) = A336158(j), for all i, j >= 1." I don't know how I would've possibly been able to transcribe these onto the desert sand.


This idea took a little longer.

Me trying to doodle my puzzle into existence instead of working on 451 homework, pt 2

However, the three characters were easy enough; if Inazuma is the land of environmental disasters and puzzles, and Sumeru is the land of fantastic story-driven content and racism, then Fontaine is the land of absolute nonsense (in the Alice in Wonderland sense). And what better way to represent that than with the three characters with really long and hard-to-pronounce names?

I did want to include Elynas somehow if at all possible, so I considered making some sort of Melusine puzzle (after all, there are a lot of Melusines). I also considered making some sort of artist puzzle with Salsa, Mamere, and Julien, but that didn't get very far. I did a whole survey of Fontaine for different puzzle mechanics (there are a surprising amount of them, even if the puzzles are all pretty easy), but nothing popped out to me as something I wanted to use. Finally, I settled on the idea of a music puzzle, since I wanted to incorporate French somehow and some of Fontaine's songs conveniently have French titles.

It was another day in ethics class when the replacing-letters idea finally came to me. By coincidence, I found that the three names share exactly three letters.

I found the actual sheet of paper lol

And thus I set out to find songs to use.

My candidate song playlist

I kind of wanted the songs to be thematic, so Neuvillette was definitely getting the rain one (and that was even before I did his story quest... I loved that track before it was cool). I also wanted to just include songs I liked for solvers to listen to, so Arlecchino got the Fontaine dusk one (since although the HotH reveal music is much more thematic, I don't like listening to it and I really wanted this track to be in there). I was considering using the Fleuve Cendre theme for Wriothesley, since I didn't know where he actually came from and this... sort of fit the vibes, I guess (and the 4.1 soundtrack isn't released). At some point, I just wrote "this subpuzzle is blocked until Oct 17" in the hopes that his story quest would give me a good song to use. However, listening to the playlist, I realized that I could sneak Elynas in there with "Le dormeur du val", which is played somewhere in their intestines or something. Now, I just had to hold out hope that, with these song choices, the puzzle would actually be constructable!

It was clear that it would be impossible to have single letters map to single notes. It was also clear, after a survey of Fontainian puzzle mechanics, that most puzzles were either Arkhe-related or involved using space/perspective (which I couldn't really use in a puzzle) (although there was a really cool clockwork puzzle too in the Fortress whose mechanic I didn't want to spoil regardless). Thus, the idea to use Pneumosia in my puzzle to map each note to two letters was born.

Pneumosia being literally everywhere

Even with two letters per note, though, construction was very constrained, simply because the three songs share so many notes. I turned to virtualpiano.net to help, transcribing the entirety of each song (besides the impossible parts) with letters and trying to identify substrings that looked not terrible.

I can't believe this is real

Finally, at some point, I got something that worked! So it was time for the artwork. I thought using the 4.1 version art would be super cool, but I had to manually redact the Traveler, Paimon, and Furina from the image.

Almost all of this was done just using the blend tool

bye bye furina

I've literally had nightmares about the blend tool between then and now

Thankfully, I thought I would never have to blend stuff into oblivion again. Then, all I had left to do was code up the dragging stuff, which was surprisingly hard and touch-unfriendly (I tried making a whole different dragging mechanism for touch screens that works on my laptop's touch emulator, but not on actual touch devices :( ). And it happened! I was pretty satisfied with the puzzle in the end; it also felt very Fontaine.

This subpuzzle in a nutshell


This one was the hardest to come up with. Sumeru's rainforest doesn't really have serious puzzles; all it has are the simple line nurseries, a set of riddles given out by an Akademia person, and of course Aranyaka/Aranara-related things, which is a spoiler for Aranyaka.

I didn't want to spoil Aranyaka, so at first, I tried to just write a riddle puzzle. It wasn't very good.

please don't even try to read this, it's embarrassing (i'm so sorry to the people who actually had to testsolve it when i had no better ideas)

It was another day in ethics class (specifically, the Tuesday before the hunt) when I came up with a second idea: although I couldn't just spoil Aranyaka, I could still make an Aranyaka-inspired puzzle. The idea in question was to make some sort of 2d game that went around Sumeru.

Some ideas

However, this seemed really hard to implement as well, so I took another look at Aranyaka material. The gadget that keeps track of the quest is split into three different sections, which could correspond nicely to three characters.

The section of Aranyaka Nahida's section was based off of (and Nahida's skill, which was the inspiration for the little window)

The section of Aranyaka Faruzan's section was based off of (and some actual writing on the wall with lines to indicate order, which the wall drawings are based off of)

The section of Aranyaka Collei's section was based off of (specifically, I based hers off of the quest "An Unwavering Culinary Dream", which is less "spoilery" than the other Aranyaka quests)

And there it was! All I had to do, then, was take a good picture of the Apam Woods and draw all the puzzly stuff into it. Unfortunately, my phone is quite literally beneath the minimum specs needed to run the game, and my resolution looks like this:

So I had to go back into my nightmares and break out the blend brush again, blending the image into not looking really grainy and blending the Genshin watermarks into oblivion.

bye bye logo

Taking the photos was pretty fun, though, especially because the area is so pretty. The scale is hard to tell from just a few pictures. Here are some of my favorites I took in Nahida's photoshoot:

It was so fun trying to get her to take different poses!

My poor camera roll...

By the time I started drawing things onto the image, it was late night the day before the hunt. At about 4pm after class, I found a seat in the UC and just didn't get up for the next ten hours or so, just drawing away.

I swear I spent an hour drawing this one starshroom

I'm not sure I'm entirely happy with the way the puzzle turned out in the end; making both the first and second subpuzzles indexing-based made cheesing the puzzle kind of easy. However, at the very least, this subpuzzle felt pretty Sumeru, and it was easy (comparatively), which is how things were meant to be (Inazuma/Khaenri'ah being hard and Sumeru/Fontaine being easy, to align with respective puzzles from their regions).

In the end

Usually when I write puzzles, I feel very sorry for the solvers in the end. This time, though, I'm not sorry at all. These puzzles felt so right, and I'm proud of them! (Besides the desert one, but I don't care too much for desert puzzles, scripts, or characters anyhow; the region itself is vastly underappreciated, but that's from more of a storyline and exploration perspective.) In fact, I think things would've been even nicer if I could've made the puzzle harder so that people couldn't just skip the whole thing by confirming trigram guesses (although not solving every subpuzzle was intended; ideally, people would just pick the types of puzzles they like and fill in the blanks).

The process was kind of frusterating to go through without having anyone to talk to about it, but that's what I took so many progress screenshots for; I wanted to get this stuff out into the void now for the sake of myself from a month ago.

Thank you for reading this far (especially Cheshire); if puzzles can be written for the single person who would really enjoy them, then these author's notes were written for you.

Hopefully, I'll stop writing terribly long and complicated puzzles and come back with nice, elegant, and shorter puzzles next semester!