The puzzle started from there. The final answer, I decided, would come from “increasing and diffusing” (that is, alphabetically incrementing and then anagramming) letters from other answers. I spent literally days trying to figure out what a settle on the perfect nine-letter answer. It had to be American but universal, meaningful but not guessable. (“Knowledge” was out, obviously.) I went through notepads full of possibilities–lots of Latin phrases about light and science and so forth, I remember–before hitting on Walt Whitman’s famous confession “I contain multitudes,” which seemed emblematic of American history and the Smithsonian collection. Even better, it had eighteen letters: two from each password. Another week went into selecting a variety of Smithsonian artifacts (again, iconic but not guessable, which was tricky) with the right letters in the right places.
The most exciting part of this process was the unlooked-for serendipities that kept cropping up. I had already thought of doing a cryptogram puzzle where the answer was hidden not in the plaintext but in the key, which I didn’t think I’d ever seen before. But this meant finding a famous Smithsonian artifact with no repeated letters in its name, which turned out to be harder than it sounded. I also knew that one of my treasures would need to have a ‘Z’ in the name, to “increase and diffuse” into the lone ‘A’ in “I CONTAIN MULTITUDES.” Bingo, “FONZ’S JACKET” solved both problems. And I liked that the elusive ‘A’ would come from the guy who always said, “Aaaaaaayy.”
Along the same lines, I wanted to do an aviation puzzle built around airport codes, and I discovered that “GLAMOROUS GLENNIS” (a) was a famous Smithsonian plane, (b) had the right letters in the right places for the final answer, and (c) could be spelled out using valid IATA codes. (It turns out that many plausible three-letter abbreviations are not used for airports.) I wanted a crossword that would have a code concealed in it even after the grid was filled in…and presto, “MORSE TELEGRAPH” had the right letters, and was 15 characters including the space (typical crossword grid size) and worked out neatly with a code-based solution. I wanted to do a spatial puzzle, but wasn’t sure how that would work online…but then realized that origami was the perfect fit for the animals in the Smithsonian’s natural history wing. The iconic “HOPE DIAMOND” was the perfect fit for a 19th-century-style riddle with oblique hints about hope and diamonds, inspired by memorizing all of Gollum and Bilbo’s riddles from a very young age. (Also by this still-unsolved Samuel Wilberforce “enigma,” which I can still recite verbatim.) “MOON ROCK” turned out to be a great choice for a space-themed logic puzzle, because all the O’s and the C looked like phases of various heavenly bodies. This took over a month to come together, but it’s hard to convey how exciting it was when something actually worked. (Or how frustrating it was when I couldn’t quite make something work. Designing a logic puzzle whose answer had to conceal “MOON ROCK” took days before I finally had the right idea.)
With all nine treasures accounted for, I waded into the actual construction of the puzzles. The biggest time commitments turned out to be the crossword and the hidden picture, for pretty much the same reason: these were both things I liked very much, but didn’t (yet) have the chops to put together myself. Crossword construction is a very unforgiving art, and I have friends who are very good at it, but I’d literally never even tried to make one. And this one was going to be a doozy: every single ‘O’ and ‘A’ in the grid had to be in the right spot, and the theme answers had to hold instructions as to how to decode the grid (again, without over- or underusing O’s and A’s). It was a real baptism by fire. (Originally there was going to be an acrostic message in the clues as well. Reader, I bailed on that idea fast.)
The hidden picture was even worse: I like to draw, but haven’t been serious about pen-and-ink in years. And yet suddently I needed to produce an intricate drawing of the Smithsonian with 21 state outlines hidden there in the proper order. (Some solvers, I believe, never realized that the left-to-right order of the hidden states was needed to spell out the answer perfectly. It wasn’t just an anagram.) Oh, and I had to write a 50-letter poem about the Smithsonian with very precise word lengths and 21 letters that fell just so, and it would be nice if it rhymed. Ugh. Both puzzles took well over a week each. I had no idea what I was doing.
But the little origami elephant, which I’d been dreading, turned out to be a cake walk. I generally suck at these kinds of spatial puzzles and had never even tried origami before, so I outsourced the job to my origami-savvy sister, explaining the basic idea. Was it even possible, I asked, to show a crease pattern and have people fold it start-to-finish with no explanatory diagrams at all? The next day she had me come over and showed me a finished pattern, including ideas on how to label stuff and where the word ‘MAMMOTH’ and all the red-herring letters would go. She’d watched a bunch of elephant-folding videos on YouTube, and prepared prototypes of three different designs. Unbelievable.
Not every puzzle turned out perfectly. I intended the airport-code puzzle to be one of the easiest of the nine, thinking that of course a list of nine aviation destinations would inspire solvers to look at IATA codes first thing. But the fake narrative I wrote to embed the code in turned out to have just too much forest for the trees, and even my very puzzle-savvy test-solvers didn’t hit on the solution right away. I intended the flight log to read as simple period pastiche, but solvers unpacked every detail of it, scouring it for clues, and were upset when some details turned out to be historically impossible. (Modern-day Namibia, it turns out, was never called “German South-West Africa” during the time period when one of the planes I mentioned was being produced. Players seized on this anachronism as Potentially Very Important, which made me feel bad. I just thought “German South-West Africa” had a cooler, more old-timey sound. I could imagine Mr. Burns from The Simpsons saying it.)
While I’m confessing to my Great American Puzzle Crimes: the last couple folds in the mammoth also weren’t labeled as consistently as they should have been, I discovered. If you looked up an origami “reverse fold,” there was really only one way to do it along the creases specified. But I made the mistake of using the “mountain” and “valley” descriptors in this step to refer to the side of the paper facing “up” to the solver, whereas in past steps they’d always referred to the printed side of the paper. (These were the only two steps where those two orientations weren’t one and the same.) Probably leaving out “mountain” and “valley” altogether on this step would have been less confusing. Ken Jennings Origami Puzzles Inc. regrets the error. (Which was mine, obviously, and not my sister’s.)
The hardest/least popular puzzle for solvers, as intended, turned out to be the presidential portrait mosaic. I knew early on I wanted a puzzle that functioned as a presidential trivia quiz, and I wanted a picture puzzle as well (using only public-domain materials where possible) and a presidential portrait mash-up seemed like a good way to accomplish both. I knew that this was going to be a slog for solvers, straining their eyes over Web versions of Smithsonian portraits for hours on end, but I thought that was okay. If the contest was really going to separate the most dedicated puzzles, not every step should be solvable by half an hour of pencil agility or five minutes of insight. At least one was going to have to be labor-intensive. This one reminded me of some old Games magazine contests (The National Scavenger Hunt, “Calculatrivia”) that were all about the research hours.
One of the contest’s top finishers, whose puzzles I normally like very much, was vehemently against the way I’d set this one up, quibbling that many of the presidential identifications don’t contribute to spelling out the final solution, which he found inelegant. I’m not sure I agree. Spelling out a message using the numbering of presidential terms only works for presidents 1-26, and it seemed a shame to leave out the most novel and recognizable presidential portraits just because they had the misfortune to come after Teddy Roosevelt. In addition, I saw that I could make the answer (“LINCOLN’S STOVEPIPE HAT”) actually take the shape of a top hat, a twist I just couldn’t say no to. But that meant there needed to be some kind of “negative space” around the hat shape. Presto, use presidents 1-26 for the hat, and president 27-44 for the background. (Since the elements from the recent presidents could appear in any order, I was free to choose cool, recognizable bits from their portraits: a vase here, a Norman Rockwell signature there. I think I played fair.) This wasn’t wasted effort, because solvers did still have to source all the picture elements. Otherwise there was no way to know (at first) which presidents were signal and which were noise.