I should start by saying that although I will be pushing back at certain places I appreciate that Jenn wrote this and don't even disagree that much, I just want to "correct the record" so to speak.

Also I am writing this fast because I'm not trying to do a polished thing, so
think of this as a textual "live reaction" post. (Please like comment and subscribe)

# Positives

lol, so, all the positives are the things that were developed first to sell the
round proposal.

I don't have anything to say here it's just really funny! Like, in brainstorming
the early pitch was "think of all the cool fractal art" we could add here, and then
Brian writes a solid metameta, and then we were like "yoooo that was fun,
okay this round idea seems real let's see what we can do"

So, you know, maybe that gives context on why the round ends so strongly. If it
didn't, the round wouldn't have made it past brainstorming. The stuff that made it
weaker were the details later.

# Negatives Round 2

I briefly talked about this elsewhere, but, yeah as the main author of The Scheme
I'm not so happy with it. So at a higher level, I'm not sure how much I'd try to
fix it. I would much rather nuke it from orbit and start from scratch.

But let's say we're going to try to fix it without much change to mechanics.

The 45 thing was just a red herring we missed. We had some feeder swaps that could
have made the total length 55. But none of us had counted the perimeter length of
The Legend, so we didn't spot the 1:1 correspondance. And then the batch testsolve
that solved The Scheme right after The Legend didn't notice the 1:1 correspondance
either. Still, not really an excuse, part of the editing / writing job is to predict
problems before testsolving occurs.

I suspect, that our post-writing decision to make The Legend be physical reinforced
that incorrect correspondence. IIRC there was exactly 1 testsolve done after we started talking to logistics about manufacturing - our last chance to catch it would have been
to say "btw this is a physical puzzle" in that testsolve. It wasn't puzzle relevant,
so we didn't. But the real solvers didn't know that.

As for splitting based on words, I distinctly remember an hours-long discussion on
the exact phrasing of the flavor text. We had testsolved a version that clued
word splitting more strongly - it got solved in about 10 minutes. Which seemed like
the wrong solve time for a meta in the AI section of the Hunt, so we spent the
next few hours debating what we could drop from the flavor text without destroying
solvability.

At the end of that debate, we decided that including "words" in the sentence
"What words can help you find your way?" was the best we could
give without fully trivializing the puzzle. Based on the testsolve data we had,
I don't think it was obvious that the eventual stacking hint would be needed.
A stuck testsolve will eventually get around to reading flavortext very carefully, and
that's what we saw in testing, with the a-ha coming at approx the 1-2 hr target
mark. (Those testsolves did not stare at The Legend which was where problems
showed up.)

This isn't to say I like the word splitting, I do think it is normally something
you clue much more strongly or don't use at all. Just got too constraint hell-ed
to find a better solution at the time for taking Wyrm feeders and was boxed in
by the "should take > 10 min" context of the Hunt it was in.

I will reiterate what I said before - The Scheme is not a fixable meta concept.
There are some small fixes to make but it is a concept that should be nuked from
orbit. We didn't, oh well.

# Negatives Round 3

The looseness of the correspondence was directly driven from the Wyrm metameta
constraints. I think no one was a fan of it but it was necessary to make it fit
and we had a hard time letting go of the FELLOWSHIP == FELLOW SHIP pun. Like how
often do you find a reasonable meta pun that also fits metameta constraints?

The indicator idea seems like a good potential fix for this.

Steps 4-6 used to not exist in the puzzle. They got added solely to get some
triangles into the puzzle in some structural way. I can imagine the version of
the puzzle where we just give the triangular sea and put the ships on it with arrows.
It seemed a little lame at the time, we could have shipped (lol) without it but
I don't think it's that big a deal in the overall puzzle to have it.

Okay I should be more precise. Steps *3-6* used to not exist. The flag icons in
step 3 were added to introduce a source of info (the flag) for step 4.
If you cut steps 4-6 I think you are supposed to cut Step 3, because only the nautical number
matters, not the ship classification letters, and you don't want solvers to
rabbit-hole on using those. So that's why it feels okay to me. The confirmer
only exists because of the work of steps 4-6. I think the confirm step
would need to look different from the existing puzzle if you strip out that
busywork.

Step 7 being index hell, yeah, it just is. We had a small hammer for digital
indexing, it testsolved poorly, so we swapped in a sledgehammer. Eh.

On the order of the ships: the ships are ordered by reading order of the extraction.
This is *not* the same as the cycle order of the index. The extract order is

USS Favorite
USS Hamilton
USS Midas
USS Underwood
USS San Francisco
USS Jackson
USS Providence

The cycle order (for what digits to index with) is

USS Favorite
USS Hamilton
USS San Francisco
USS Midas
USS Underwood
USS Providence
USS Jackson

(Yes some testsolvers were grumpy about this.)

Swapping the order of ships breaks the puzzle, they need to be in that order.

It's possible this meta concept was just too constrained to begin with, but, as
the first submeta to get written, it felt like we had the room to afford. The round
template I looked at the most during writing was the Sci-Fi round from Mystery Hunt 2018,
and I'd say Lost at Sea == Blue Sun 6V4-178-B31 Trace Compression Block from that round.
(The meta about Firefly.) Like, reread that puzzle and solution. No disrespect,
it's pretty jank? Of the Sci-Fi metas I think it puts the strongest constraints
on its feeders. It is also eerily similar in construction to Lost at Sea - do semantic pairing
between feeder and actor, bounce through some parts of the dataset, end by doing
an unconventional extraction to get your answer.

# Interconnectedness

I don't think it's fair to compare Wyrm to Indiana Jones. The Indiana Jones round:

* Is based on a character from a well known IP, hence it has low hanging fruit
to reference for resonance. MH2023 needed to build up Wyrm during Hunt.
* Doesn't have constraints on the answers to their metas (as far as I can tell they don't
matter at all)

As mentioned earlier I think the 2018 Sci-Fi round is a better comparison. In that
round, the metameta places constraints on both the meta answers and feeder answers,
and the feeders were shared between submetas. I recognize that Indiana Jones uses
paired answers, and so it makes sense to compare on that axis, but the answer constraints
from Sci-Fi were much more similar to our struggles at construction time.

There was probably room to play up Wyrm's love of triangles from the creative side
(there's a reason Wyrm's background art is triangular graph paper). I think creative
was just busy on doing so.

In general though - yes writing Wyrm metas was a lot. I wrote about this earlier
but basically we were in constraint hell, and then we infected Museum with a bit
of our constraint hell when we decided to repeat feeders. There was a very
deliberate planning discussion, about how much we could allow the Wyrm metameta to
pollute the quality of its feeder metas, and our conclusion was that if we wanted
our round vision to work we'd have to live with that. I remember a rough target of
"2 good metas, 2 meh metas" as the best we could do. I think we hit that
where people mostly liked Legend + Collage and did not like the rest.

WHO WOULD WIN?

teammate authors spending many hours brainstorming good meta ideas

OR

one loopy AI

(the loopy AI won)

# Collage

Yeah this definitely has Fridge Logic. Just didn't come up with a good way to
make it not be Fridge Logic. I believe, that the platonic ideal of "a metapuzzle
solvable from 0 feeders that you backsolve later" is the Hall of Innovation meta.
It's super cool! We were not going to be able to pull that trick twice, I
think it's a bit miraculous Innovation exists at all.

# Partial Progress

This did get brought up in editor feedback, that the round did not have much
exciting to it if your team did not make it to the looping step. The proposed fix for
that was to repeat feeders from the Museum. As I've mentioned earlier, I think
this caused some other problems and wasn't sufficiently "cool" as a hook relative
to the difficulties it introduced. (But maybe this is wrong? If it's enough
of a basis to inspire an entire alternate round concept, maybe it is compelling...)

My assumption with most Mystery Hunts is that many teams do not see the whole thing,
but people will usually open the archives afterwards, and so even if they did
not experience it live, they could at least appreciate the structure afterwards
(if they're into that sort of thing).

# Fixes

Given that I think our decision to plagiarize Museum puzzles was a mistake...it
s is kinda wild to read a round proposal about doing it more?

That being said my first reaction was "this shouldn't work" and my second was
"nevermind it should". I think it reads as more constrained than it actually is,
since you are allowed to change the shells of the feeders to extract what you want
pre and post-copy.

I will say, that I don't think we would have ever come up with this idea in early
2022. Good generative AI didn't appear until April 2022 for DALL-E 2 and December 2022 for
ChatGPT, so that storytelling angle wasn't going to exist during Hunt writing.
Wyrm was 100% "do some loopy shit" in my mind. The gimmick of reusing feeders
did not feel like a "Wyrm-sourced" gimmick to me. It felt like a Hunt-wide gimmick
to bridge the Factory and AI rounds, where Wyrm happened to be the bridge.

I also suspect, that if the more plagiarism based round proposal was sent, editors would have said "isn't this just Quest Coast from last year?" and asked for changes.

# Ending Thoughts

So, reiterating the start - yeah I agree with a lot of the criticism around
Wyrm's structure. I'd say the key is "flawed in an interesting way". It does
cool stuff and the flaws are because of what that forces elsewhere. I've read
similar criticism about GPH 2022's structure. A lot of the motivation for the
AI rounds was "let's try some wacky stuff" and I don't think it's so bad if
not all of it landed for people. I've seen enough "Wyrm is my favorite", "Eye is
my favorite", etc. to believe that we at least hit the goal of writing weird
enough things that each round could find its fans, even if it did not attain
broad appeal.