How Crossword Puzzles Are Really Made

Editor's Note: Last week, we published an item on how crossword puzzles are made. As many, many readers pointed out, we didn't have our facts straight. You deserve better. So today we've enlisted the help of professional crossword puzzle writer Matt Gaffney. Matt currently creates crosswords for The Week, New York Magazine, and  Washingtonian, and he also wrote our book of crosswords. Let's try this again.


If you’d like to solve the puzzle discussed in this article before reading further, click here.

When I tell people at parties that I write crossword puzzles for a living, by far the most common question they ask is, “Would you excuse me for a moment?” If that level of social esteem sounds like something to which you might aspire, follow along as I construct a crossword puzzle from start to finish.

OK, the real first question they ask at parties is, “Which do you write first, the clues or the grid?” Answer: neither. The first thing you do is come up with your puzzle’s theme.

A crossword puzzle’s theme is some unifying motif among certain answers in the grid, generally the longest answers. Crossword themes range from vaguely comprehensible to hellishly complex. For this puzzle we’ll stick to the former.

With an eye toward impressing the client, I decide right off the bat that MENTAL FLOSS should be one of this puzzle’s theme entries, and therefore search for a way to incorporate that phrase into a wider, hopefully amusing pattern. A few minutes of theme-storming later and I spot the TV show ALF bridging the two words in "MENT AL FLOSS." That’s kind of amusing, and my spidey-sense tells me there could be something theme-worthy here.

I decide that this puzzle’s theme will feature two-word phrases bridged by a one-word TV show title. Hidden words are a common theme gimmick, and if I can come up with a fitting title that explains the wordplay involved—something like “We Interrupt This Broadcast,” playing on how the show is “interrupted” by the break between the two words—then we’ll be golden. And I’ll put circles in the grid to highlight the shows, since it’d be tough to understand the theme without them.

Time to make a list of all the one-word TV shows I can find. Google and Wikipedia are my friends here, but my BFF today is IMDb, specifically this page. After 30 minutes of list-scouring, I’ve scribbled down the best few dozen candidate shows, so it’s time to look for the ones that can hide in a two-word phrase. Three-letter shows like ALF or CSI (in "TRAFFI C SIGNS," say) are fine, but it’d be much cooler to use four-, five-, or even a six-letter show if I can find some that work. The phrases I settle on need to be symmetrical in length so they can offset each other in the grid. So if I'm going to use the 11-letter MENTAL FLOSS in the top left, I'll also need an 11-letter entry in the bottom right as its mirror image. The only exception is an entry going across the middle of the grid, which provides its own symmetry.

When I’m done paring the list down, my theme entries look like this:

MENT AL FLOSS hiding “ALF” (11 letters)

JOE L OSTEEN hiding “Lost” (10 letters)

ALABA MA SHAKES hiding “M*A*S*H”(13 letters)

S WING STATE hiding “Wings” (10 letters)

LI VE EPISODE hiding “Veep” (11 letters)

So that’s a decent set, and I like the mix of recent and old-school programming so everyone from teenagers to Baby Boomers and beyond can relate.

I didn’t get my six-letter show, but had a few close calls: COM BAT MANEUVER isn’t quite strong enough (just 75k Google hits) for “Batman,” and “Roswell” came within one letter of being a killer seven-letter show hidden in HE RO’S WELCOME. Can’t get everything you want in life, but this is still a B+ group. Did I miss any good ones? Let me know in the comments if so.

Black square placement is next, guided by some long-established rules: Entries must be at least three letters long, so no TV Guide-style two letter words allowed. The grid must have “180-degree rotational symmetry,” meaning that if you turn it upside-down, the pattern of black squares will remain the same. And finally, the maximum word count permitted in the grid is 78.

With theme entries and black squares placed, my grid looks like this:

Now for the toughest part of the process: Filling the grid, and not just with words that fit but with ones that are fun as well. "AREA" and "ARIA" and "ONO" and "ERIE" are all perfectly good words, but they also appear so often in crosswords that it’s tough to come up with an original clue. You know how many times I’ve scoured Yoko Ono’s Wikipedia page looking for something new to say about her? A lot of times.

Another common party question: “Can’t a computer just fill the grid for you?” Answer: Yes, sort of. A few holdouts (like myself, about 25% of the time) still use graph paper and pencil to fill their grids, which has its own advantages (your brain has fresh entries that haven’t made it into databases yet) and disadvantages (it takes longer, and you might miss something obvious the computer finds in two seconds), but mostly, the process is semi-computerized now, a dance between the human brain’s judgment and the silicon monster’s brute force. You still need both to get a good final product, but autofill programs are gradually making the human contribution less and less important.

I start my fill-work in the center of the grid, a common divide-and-conquer strategy which will leave me with the more blocked-off edges to knock out later. If I start at the edges, I might later find it impossible to finish the center that connects them all, and then I’d be back to square one (rim shot).

The center is also a logical place to begin here since there are three theme entries running through it, which constrains my fill choices. The peripheral sections all have just one or two themers in the area, which makes life easier for me out there.

After about two hours I’ve got a fill I’m satisfied with, especially considering the five theme entries (which is on the high end). As with the theme entries, the fill has a decent balance of contemporary ("HBO," "OMG," "GOLD’S" Gym) and classical ("KABUL," "PARIS," "LBJ"). I'm not too thrilled with having to use the obscure "CIS," the abbreviation "ONT," or the partial "DEO," but that’s how it has to be with crosswords, and I've kept the dreck to a minimum. Plus there’s some catchy stuff in there, like "EASY NOW," "CON JOB," and the cross-referenced "STEELY / DAN" to balance it out.

Now it’s time to clue this thing. I aim for that same mix of old and new, with a few misdirecting clues thrown in to keep the solver on notice. My favorites are [Word that reverses to a “Seinfeld” character] for "REMARK" and [Felt something on your head?] for "FEDORA."

Sometimes, a marginal entry can be rescued with a good clue, like ONT here. Abbreviations are considered suboptimal, but this one is saved by the misdirecting [Neighbor of N.Y. and Minn]. Hopefully you were trying to envision what U.S. state could possibly border both of those.

Finally, it’s time to fact-check all my clues. Crossword editing tip: check everything, especially the clues you know something about already. That’s where the errors come. For instance, I’ve read three or four Michener books, but "IBERIA" at 2-Down isn’t one of them. But a little knowledge is a dangerous thing; my original clue referred to it as a Michener “novel,” since all of his books that I’ve read have been novels. I checked this one to be safe—and sure enough, Iberia is a work of non-fiction.

The very last step is a quick test-solve to make sure I didn’t miss anything. It all looks good, so I export the puzzle into a tidy e-file and send it off to my editor at 17-Across (which rhymes with its answer, "MENTAL FLOSS").

Total time elapsed, from theme concept to clicking “send,” is a little under six hours. And here she is:

Original image
iStock // Ekaterina Minaeva
Man Buys Two Metric Tons of LEGO Bricks; Sorts Them Via Machine Learning
May 21, 2017
Original image
iStock // Ekaterina Minaeva

Jacques Mattheij made a small, but awesome, mistake. He went on eBay one evening and bid on a bunch of bulk LEGO brick auctions, then went to sleep. Upon waking, he discovered that he was the high bidder on many, and was now the proud owner of two tons of LEGO bricks. (This is about 4400 pounds.) He wrote, "[L]esson 1: if you win almost all bids you are bidding too high."

Mattheij had noticed that bulk, unsorted bricks sell for something like €10/kilogram, whereas sets are roughly €40/kg and rare parts go for up to €100/kg. Much of the value of the bricks is in their sorting. If he could reduce the entropy of these bins of unsorted bricks, he could make a tidy profit. While many people do this work by hand, the problem is enormous—just the kind of challenge for a computer. Mattheij writes:

There are 38000+ shapes and there are 100+ possible shades of color (you can roughly tell how old someone is by asking them what lego colors they remember from their youth).

In the following months, Mattheij built a proof-of-concept sorting system using, of course, LEGO. He broke the problem down into a series of sub-problems (including "feeding LEGO reliably from a hopper is surprisingly hard," one of those facts of nature that will stymie even the best system design). After tinkering with the prototype at length, he expanded the system to a surprisingly complex system of conveyer belts (powered by a home treadmill), various pieces of cabinetry, and "copious quantities of crazy glue."

Here's a video showing the current system running at low speed:

The key part of the system was running the bricks past a camera paired with a computer running a neural net-based image classifier. That allows the computer (when sufficiently trained on brick images) to recognize bricks and thus categorize them by color, shape, or other parameters. Remember that as bricks pass by, they can be in any orientation, can be dirty, can even be stuck to other pieces. So having a flexible software system is key to recognizing—in a fraction of a second—what a given brick is, in order to sort it out. When a match is found, a jet of compressed air pops the piece off the conveyer belt and into a waiting bin.

After much experimentation, Mattheij rewrote the software (several times in fact) to accomplish a variety of basic tasks. At its core, the system takes images from a webcam and feeds them to a neural network to do the classification. Of course, the neural net needs to be "trained" by showing it lots of images, and telling it what those images represent. Mattheij's breakthrough was allowing the machine to effectively train itself, with guidance: Running pieces through allows the system to take its own photos, make a guess, and build on that guess. As long as Mattheij corrects the incorrect guesses, he ends up with a decent (and self-reinforcing) corpus of training data. As the machine continues running, it can rack up more training, allowing it to recognize a broad variety of pieces on the fly.

Here's another video, focusing on how the pieces move on conveyer belts (running at slow speed so puny humans can follow). You can also see the air jets in action:

In an email interview, Mattheij told Mental Floss that the system currently sorts LEGO bricks into more than 50 categories. It can also be run in a color-sorting mode to bin the parts across 12 color groups. (Thus at present you'd likely do a two-pass sort on the bricks: once for shape, then a separate pass for color.) He continues to refine the system, with a focus on making its recognition abilities faster. At some point down the line, he plans to make the software portion open source. You're on your own as far as building conveyer belts, bins, and so forth.

Check out Mattheij's writeup in two parts for more information. It starts with an overview of the story, followed up with a deep dive on the software. He's also tweeting about the project (among other things). And if you look around a bit, you'll find bulk LEGO brick auctions online—it's definitely a thing!

Original image
Sponsor Content: BarkBox
8 Common Dog Behaviors, Decoded
May 25, 2017
Original image

Dogs are a lot more complicated than we give them credit for. As a result, sometimes things get lost in translation. We’ve yet to invent a dog-to-English translator, but there are certain behaviors you can learn to read in order to better understand what your dog is trying to tell you. The more tuned-in you are to your dog’s emotions, the better you’ll be able to respond—whether that means giving her some space or welcoming a wet, slobbery kiss. 

1. What you’ll see: Your dog is standing with his legs and body relaxed and tail low. His ears are up, but not pointed forward. His mouth is slightly open, he’s panting lightly, and his tongue is loose. His eyes? Soft or maybe slightly squinty from getting his smile on.

What it means: “Hey there, friend!” Your pup is in a calm, relaxed state. He’s open to mingling, which means you can feel comfortable letting friends say hi.

2. What you’ll see: Your dog is standing with her body leaning forward. Her ears are erect and angled forward—or have at least perked up if they’re floppy—and her mouth is closed. Her tail might be sticking out horizontally or sticking straight up and wagging slightly.

What it means: “Hark! Who goes there?!” Something caught your pup’s attention and now she’s on high alert, trying to discern whether or not the person, animal, or situation is a threat. She’ll likely stay on guard until she feels safe or becomes distracted.

3. What you’ll see: Your dog is standing, leaning slightly forward. His body and legs are tense, and his hackles—those hairs along his back and neck—are raised. His tail is stiff and twitching, not swooping playfully. His mouth is open, teeth are exposed, and he may be snarling, snapping, or barking excessively.

What it means: “Don’t mess with me!” This dog is asserting his social dominance and letting others know that he might attack if they don’t defer accordingly. A dog in this stance could be either offensively aggressive or defensively aggressive. If you encounter a dog in this state, play it safe and back away slowly without making eye contact.

4. What you’ll see: As another dog approaches, your dog lies down on his back with his tail tucked in between his legs. His paws are tucked in too, his ears are flat, and he isn’t making direct eye contact with the other dog standing over him.

What it means: “I come in peace!” Your pooch is displaying signs of submission to a more dominant dog, conveying total surrender to avoid physical confrontation. Other, less obvious, signs of submission include ears that are flattened back against the head, an avoidance of eye contact, a tongue flick, and bared teeth. Yup—a dog might bare his teeth while still being submissive, but they’ll likely be clenched together, the lips opened horizontally rather than curled up to show the front canines. A submissive dog will also slink backward or inward rather than forward, which would indicate more aggressive behavior.

5. What you’ll see: Your dog is crouching with her back hunched, tail tucked, and the corner of her mouth pulled back with lips slightly curled. Her shoulders, or hackles, are raised and her ears are flattened. She’s avoiding eye contact.

What it means: “I’m scared, but will fight you if I have to.” This dog’s fight or flight instincts have been activated. It’s best to keep your distance from a dog in this emotional state because she could attack if she feels cornered.

6. What you’ll see: You’re staring at your dog, holding eye contact. Your dog looks away from you, tentatively looks back, then looks away again. After some time, he licks his chops and yawns.

What it means: “I don’t know what’s going on and it’s weirding me out.” Your dog doesn’t know what to make of the situation, but rather than nipping or barking, he’ll stick to behaviors he knows are OK, like yawning, licking his chops, or shaking as if he’s wet. You’ll want to intervene by removing whatever it is causing him discomfort—such as an overly grabby child—and giving him some space to relax.

7. What you’ll see: Your dog has her front paws bent and lowered onto the ground with her rear in the air. Her body is relaxed, loose, and wiggly, and her tail is up and wagging from side to side. She might also let out a high-pitched or impatient bark.

What it means: “What’s the hold up? Let’s play!” This classic stance, known to dog trainers and behaviorists as “the play bow,” is a sign she’s ready to let the good times roll. Get ready for a round of fetch or tug of war, or for a good long outing at the dog park.

8. What you’ll see: You’ve just gotten home from work and your dog rushes over. He can’t stop wiggling his backside, and he may even lower himself into a giant stretch, like he’s doing yoga.

What it means: “OhmygoshImsohappytoseeyou I love you so much you’re my best friend foreverandeverandever!!!!” This one’s easy: Your pup is overjoyed his BFF is back. That big stretch is something dogs don’t pull out for just anyone; they save that for the people they truly love. Show him you feel the same way with a good belly rub and a handful of his favorite treats.

The best way to say “I love you” in dog? A monthly subscription to BarkBox. Your favorite pup will get a package filled with treats, toys, and other good stuff (and in return, you’ll probably get lots of sloppy kisses). Visit BarkBox to learn more.