New York Times, Tuesday, April 9, 2019

This puzzle was published pretty quickly. It's been less than a year since I wrote it, and only about nine months since I got the always thrilling "Crosswords — Yes" email from the editorial crew. Over time, I've noticed that puzzles that depart from common theme types often make it into print much faster than more traditional grids. So if you're looking for speedy publication, my advice is to push the boundaries of originality and novelty!

To build the grid, I started by finding and placing some "long" across entries into a promising configuration. There aren't too many valid English semordnilaps (words that have different meanings when read forwards and backward) with more than five letters, so my options were pretty limited.

Next, I slowly, painfully filled in each section while referencing two different word lists — my normal list and a special one that only contained semordnilaps. As is typical with highly constrained stunt puzzles, I frequently reached dead ends and had to tear up what I'd already created. While filling, I also had to avoid words whose reverse I'd already put somewhere else.

An unfortunate side effect of the paucity of longer semordnilaps was that I had to keep most of the across slots at five letters or fewer. I wanted to incorporate more longer semordnilaps, but it was too constraining on the rest of the fill.

I'm not sure I've ever seen a published NYT puzzle with only ten entries longer than five letters. It would be interesting to know if that's a record. Likewise, XWord Info only tracks puzzles with the longest average word length– I'd be curious to know where this one stacks up against others with a very low average word length.

Whoa! I've seen tons of puzzles with backward entries. Puzzles with dual clues. Puzzles with all the across answers having something in common. Puzzles with words reading one thing one way and another thing another way.

But I've never seen anything quite like this. Elements of all of the above stitched together brilliantly.

Such a clever idea to give the solver two clues for each across entry, leaving it up to them to figure out which applies in the forward direction, and which in the reverse. I've seen most of these "emordnilap" words before (emordnilap = palindrome backward), but the notion of clueing both the regular word and its emordnilap is a great out-of-the-box idea.

My solve was much slower than for a regular Tuesday, and my enjoyment flagged through the middle, as the trick got a bit old. But after finishing, I had to sit back and admire the concept and construction. So, so, so difficult to get every single across answer to work this way.

There were plenty of gluey spots, not just ANART EPT SSS REPUT DIALLED, but backward stuff like RETAR. The overall impact was so strong that I was easily able to brush those aside. Heavy crossword glue in the service of a great theme is fine by me.

It's so rare that a puzzle stands out as something entirely new. This is one of them.

Alex was curious about how this stacks up (down?) to other puzzles with low average word lengths. It just makes the top 10.

I quickly put together a thumbnail view sorted by word length. It's a little hard to see, but you can get the idea.