Back to Wrap-Up

Questions have been edited slightly for clarity.

Hunt Structure / Design

Was the hunt intended to be harder than last year's, or was that a natural byproduct of a more experienced writing team? Was it harder than you expected it to be?

A more experienced writing team doesn't necessarily mean a harder hunt, but it does seem to be a trend. We set out to create puzzles of similar difficulty to last year, and based on solve stats this looks true. The main difference was entanglement layering additional difficulty and length.

Were there any major changes made to the hunt after the full hunt testsolves?

We changed the unlock structure after the 1st full hunt testsolve. Previously, unlocks were balanced between Matt and Emma, and both The Mystical Plaza and The Sword Swallowers were at the end of the unlock track. Testsolvers found it demoralizing to not get a meta unlock for a long time, so we rearranged unlock order to move up first meta unlock.

Entanglement Structure

What was the process behind constructing the linked puzzles? They're so involved that I imagine you would have to have had teams of people work on each one.

Not all of them were constructed by large teams. However, all of them did go through lots of discussion and iteration over what information is being used by which puzzle, whether there really are two independent puzzles, and whether the entanglement is discoverable. Often (but not always), we had these discussions before constructing even a prototype.

General Puzzle Writing

Any advice for helping to ensure the hunt-writing process goes smoothly, consistently, and steadily? (emphasis on the last two ummm lol)

We've yet to find anything better than deadlines. We held weekly meetings, which helped give a cadence to hunt construction (it was easiest to testsolve a puzzle if a draft was ready prior to that meeting). The batch testsolves helped provide other intermediate deadlines. The challenge is making the deadline accountable - we don't have a great answer for how to do that.

How did you work to avoid tedium in the puzzles? How did you know a puzzle was "just right" in its difficulty and solve path? How does one make a flavor text that clues enough but not too much?

Testsolving! Constructors often have an intuition on what makes a puzzle feel interesting vs tedious, and testsolving is where those intuitions are tested. Our testsolving feedback form emphasizes both fun and difficulty of the puzzle, and the authors iterate to try to improve fun while hitting their target difficulty.


How did you go about designing, writing, constructing, and coding interactive puzzles? Guiding principles for said interactive puzzles, if any?

Interactive puzzles are tricky to get right; just like any game or web app, the process is quite iterative and requires a lot of playtesting and finetuning.

Typically, we start testsolving early with “paper prototypes”, or low-fidelity versions that showcase the core mechanics. For Remember to Hydrate, this was a text-only version that described the plant state in words; for Mystery Manor, we had placeholder assets and “programmer art” or just text in place of missing images. (Last year, when testing Kyoryuful Boyfriend, we even created a fake dinosaur Discord profile and had someone type out scripted interactions!)

One important design principle is simplicity and teaching through trial rather than tutorials. Controls should always start simple and grow intuitively; the player should never be overwhelmed when just starting out. In Remember to Hydrate, we started with watering, weather, and rain, and gated the complex interactions behind plant progress. In Beep Arcade, we locked the more difficult rooms behind heart gates and introduced sections that required diagonal dashes or in order to teach these mechanics to new players (a common theme in the base game, Celeste, which has amazing level design!)

(Note that many of our “interactive” puzzles were originally completely static, including 🔗, The Bazaar, and Souvenir Photos. The interactive steps were iteratively added post-testsolve in order to aid the group solving experience.)


What was your process for all of the tiny details?

Thanks for noticing! A lot of effort went into little details like swapping out favicons and story images, hiding Matt & Emma names, and tuning the color scheme. The site uses base templates with CSS variables defined differently per dimension, which is stored in a cookie shared by every page. As for tiny details, we did QA by asking everyone to go through the site and report bugs to fix until they had trouble finding new ones.

How did synchronizing puzzles work?

We used several different forms of state throughout the hunt.

Team-wide state was shared across all users logged into the same team. This was saved in a row in the database and broadcast to users via websockets. For example, answer submissions, custom puzzle submissions (Guessing Game, DC Meet & Greet, Oxford Fiesta, etc.), Beep Arcade collectibles (stars), collage progress (Souvenir Photos), and global plant state (Remember to Hydrate) were all synced in the database.

Individual state varied depending on the use case. For each browser session, we generated a unique identifier (UUID) which was sent to the server. This helped key interactions (Pin the Tail murders, Mystery Manor location and dialogue responses) scoped to a single user. Finally, we used cookies to save the user’s last dimension, and localStorage and IndexedDB for local state including the last story notification viewed, volume setting, and more.

Your website was the nicest looking website I have seen!!! How do you even design a site like that?

Chris: Thank you! Figma was instrumental in helping us iterate quickly and maintain a basic design library (type styling, colors, components, etc.) so we could apply a fairly consistent ruleset across pages, and make wide-scale changes as needed. Aside from prototyping in Figma, for smaller and more complex pieces (like the icon effects or the card toggle), we coded working offline prototypes to make the design-to-dev handoff a bit easier.

Was it just me or does the C in Carnival Conundrum looks awfully like G?



How much work (person-hour/percentage) was put into art?

Jacqui: For each puzzle icon the artist needed to draw the line art, add the gold accents, and then color it in several times. It took roughly 2-4 hours for one of us to create all the images needed for one puzzle icon, with the meta icons taking a bit longer. So that’s maybe 150 cumulative hours for drawing the final icons, plus some time to draw the map and arrange the icons on it. But before we even got to this point, we spent a lot of time experimenting with art styles by drawing test icons that weren’t ultimately used.

Rachel: When it comes to design/ideation + illustration time, each puzzle endcap took anywhere from around 10 minutes on the low end to 3 hours on the high end for each of 43 unique endcaps (plus variants, when Matt & Emma disappear). The larger meta story cards took anywhere from 5 - 40 hours (unfortunately, I do not work quickly!), and there were 8 of those. If I had to sum a guess, I probably spent somewhere around 200 hours working on story art alone. As a percentage of total work, it’s hard to say, but everybody who worked on art was also very heavily involved in writing puzzles.

We noticed the change to our team profile pic! At what point do you suspect a team might finish and start designing/drawing their profile pics? What went into changing all the team photos, all that.

Rachel: "I want to do custom profile pictures again"

Everyone else: "Really? Wasn't that a lot of work last year?"

Rachel, having already edited 5 pictures: "did I stutter"

(During hunt, we can see the progress of all teams, so we can tell when a team will finish and prepare a profile pic before they do. Luckily the edits take less time than our puzzles.)

Specifically, what did you do to create the cursor effect on the puzzle home page?

Chris: This was fun to figure out! Each puzzle icon in the main round has three layers at any given time: the base image, the overlay (either gold strokes or the full recolor), and a mask image applied to the overlay. Using CSS mask properties, we’re able to set a really big mask image on each image (basically it’s a PNG of a black-to-transparent radial gradient, where the transparency value of each pixel gets applied to the image it’s masking). By mapping the cursor position to mask-position, offset by icon position and half the mask size, it creates the illusion of one big light source when we’re technically updating the position of multiple masks at the same time.

Alex: Fun fact - The proof of concept used jQuery. We spent many hours on a non-jQuery solution, but kept hitting issues where the mask offset was computed incorrectly. Eventually, we gave up and just added jQuery to our site. Worth it!

How did you come to the idea that the story should only be told through these comic panels (which are beautiful by the way!)?

Jacqui: One inspiration was MH 2020’s “Penny Park History” videos. I love how these each (almost) wordlessly conveyed a clear story beat. Our story seemed simple enough for a similar approach. We briefly considered doing animations ourselves, but our team’s skill sets were a better match for still comic panels, and they fit our overall tone a bit more.

Timeline / Effort

How long / much effort did it take to put all of this together?

Last year, we estimated somewhere between 5k-10k person-hours. We had more people, art, and puzzles this year, so we'll guess 7k-12k person-hours.

General Philosophy

Do you think puzzlehunts are designed to be (/ should be) completed in a weekend? Do you think less experienced teams should have more members so they can complete a puzzlehunt in the weekend, with less hints? I feel like the "long tail" experience of non-top teams is not ideal, as people go back to school/work, and then slowly make progress with hints, all while getting sick of puzzling for so long. I wonder how we can make this experience better? We tried really hard to finish by sunday but we still couldn't make it happen, even with a slightly illegal teamsize (well, not all of us have as much time/experience).

Alex Irpan: We definitely don't want top teams to spend more than a weekend, but how many finishers we want in the 1st weekend is much less clear, and I'd be surprised if we agreed on this as a team. Personally, because teammate has the ability to create hard puzzles with satisfying solves, I'd like it if teammate continues to do so. We do think about the solve experience of less experienced teams - in every hunt the majority of teams only do intro round, and we really try to make sure that's still a fun experience. Puzzlehunts are supposed to be fun. If you aren't having fun or have other obligations to your time, it's totally okay to stop.


Were there ever things you wanted to do that you couldn’t because you didn’t have the abilities? For example, you could do things some other hunts couldn’t do because you have a big web and art team (which is awesome!), so was there anything holding you back?

Alex Irpan: I think there were few things we didn't do because of abilities, and many things we didn't do because of time. If people believe in a puzzle concept they'll often work on it even if they have to learn things along the way - the main blocker is the time it'll take to learn the requisite skills to construct.

How diverse is your team?

Ages: mostly early to late 20s
Education: too many CS people
Pronouns: 64% he/him, 28% she/her, 12% they/them.

For Fun

Who are some villains that you love to hate? (Or antagonists)

Alex Irpan: Lord Tirek, for both almost winning and then losing thanks entirely to his value system. Walter White late in Breaking Bad for being an asshole. Bonesaw from Worm for general nightmares.

What is your single most favourite Olympic mascot?

Amik. It was made before the pride flag became popular, so now it's retroactively the gay beaver and that's wonderful.

Puzzle-specific Questions

[Radio Noise] Which is /your/ favorite Misaka?

Nishant: My favorite Misaka is the Original, Misaka Mikoto. She’s a very cool character, and has a lot of development, both in the main series and in her spinoff. But I also like all of the Sisters, they’re very cute and funny.

[SPPWNGP] Did you contact Cracking the Cryptic for the pumpkin patch puzzle? I thought they would have been more than glad to provide an actual video of Mark or Simon trying to solve a vague puzzle…

Patrick: We did not contact them. We also think it would be asking too much given the timeline for the puzzle, the time spent writing the transcript, recording and video editing (~5-6hrs each), and number of puzzle iterations (4).

[All That’s Left to Do is Extract] Are there any red herrings in have-you-tried?

The original draft of the Have You Tried page actually included a lot more silly suggestions; these were ultimately cut so that all of the NATO letters would be on adjacent lines. We did keep a few silly suggestions for self-referential fun (such as the nod to Tateroids, which would later appear in the puzzle). However, we didn't include any intentional red herrings. All of the prior sections are intended to be actually useful, and we hope that teams were able to use them to unblock themselves!

[Plagiarism] When was the 2021 Plagiarism puzzle written, and was it something else earlier?

Alex: In the first draft the 2021 Plagiarism was from Puzzlehunt CMU, but the plan was always to change it to a Huntinality puzzle. I picked The Dating Sim for the challenge. I also figured it was one of the more memorable puzzles. Interestingly, at the time, the CMU puzzle was almost impossible to find unless you searched "puzzlehunt", one word.

[Time of Your Life] This is one of those puzzles that are "easier said than done"; did the author themselves make the graphics as well? Were the graphics ready before testsolving? Did it change a lot during writing?

Jacqui: Yes, I made the graphics for this puzzle. While I was playing around with what the writing system should look like, I used PowerPoint shapes and animations. The first testsolve of the puzzle was done using videos I exported from PowerPoint. When I started making changes for the second testsolve, I switched to animated SVGs. I spent a few weekends in my armchair just writing the SVG code for all the animations—some of the most pleasant puzzle-construction work I’ve ever done. There were a few more iterations after this, but fortunately nothing too significant changed.

[Unhealthy] What is in the second floor of the warehouse and the basement of the shack in Unhealthy? Is that where the teapot is? Was Unhealthy inspired from GPH 2019 "Unsafe"?

Liam: Construction, probably. (You cannot visit those areas in the game.) The teapot is elsewhere on the site. Unhealthy was absolutely inspired by Unsafe, one of my favorite puzzles of all time. I tried to avoid hewing too close to Unsafe, but parts of it, including the tone at certain points in the narration and the deaths, were definite inspirations.

[DCMaG] Did any of the DC Meet and Greet puzzles change after QoDE? What was /your/ favorite DC Meet and Greet minipuzzle?

Our favorites are answered in authors notes. Nothing changed after QoDE besides fixing various errata, mostly related to Odd Man.

[Single Elim] Single Elimination allowed a bunch of nonsense two-letter "words". Did you intentionally use this wordlist to make the puzzle easier?

We did intentionally choose a rather large wordlist (SCOWL at default size, with all regional variants), though we didn’t think about 2-letter words in particular.

[Hydrate/MM] Is Mystery Manor and Kyoryuful Boyfriend related in any way?

Mystery Manor is the sequel to Kyoryuful Boyfriend. It takes place 400 years in the future when all the dinosaurs have evolved into birds.

[Hall of Mirrors] For the final meta, did you fix all the answers ahead of time, or let authors choose any pair that fit the constraints? Which other answer pairs have you considered for the final meta?

We fixed all the answers ahead of time - while it would have been nice to have a degree of flexibility, ultimately we needed really strong pairs to definitely clue the mirror transformations, and we didn’t want to run into an unfortunately weak pair halfway through hunt construction. Our favorite unused answer pair, which wasn’t used because it didn’t fit the grid and absolutely would have been included if it did, was VENTI/LONG - see if you can figure out the mirrors needed for that one.

WHAT IS THE TEAPOT?? (and the story behind it)

Last year, we had a puzzle about HTTP status codes that culminated in finding a teapot. We planned to reimplement all the easter eggs, but due to other priorities we only got to the 404 page and teapot page. Based on the feedback form, only 12% of solvers found the teapot, so you're not alone.