New York Times, Thursday, September 7, 2017

The key to constructing this puzzle was finding workable downs for the left-hand side of the grid. Having the acrosses in alphabetical order requires that those words (ABC, CHILL, and MOSSY) monotonically increase in alphabetosity (uh, that's a thing...trust me).

With those three words chosen, I used manual trial and error to select workable alphabetically correct starts for each of the acrosses (i.e., fixing in place one to three of the starting letters for each word) while also coming up with a grid structure that supported some good fill. Similar to the left-hand side of the grid, deploying contiguous vertical blocks (i.e., the ones next to 8-Down and 60-Down) means that the neighboring letters must be in alphabetical order (i.e., the AB of ABLUSH and all of STY).

The Utah block at the top of the grid enables relative freedom for the NE corner in light of the constraint of having ABC at 1-Down. Without those two cheater squares, the neighboring down word would need to start with ABC, ACC, BBC, or BCC (if no cheater square is used) or BC or CC (with a single cheater square). Instead, I just needed a word that starts with AB (a much easier constraint to fill around). This freedom was especially important because the black square under 7-Down (REC) was fixed in place by its symmetry to the one above 60-Down (STY).

Interestingly, my first iteration of an alphabetical order puzzle had 78 words including three across answers in each row at the top and bottom of the grid, which made construction very difficult. The resulting grid strain also made the fill pretty blah. Going down to 76 words (and more importantly reducing the number of words in the top and bottom rows) made things both easier and better (a rare combination!).

Ah, that feeling of having given up on an idea years ago, deeming it impossible … only to see another constructor make it happen. Impressive construction today, Alex putting all the across answers in alphabetical, or ABC, order.

I had been working with a grid containing exactly 26 across answers, one for each letter of the alphabet, but no matter how long I flailed away, it just wasn't happening. I like Alex's approach, which is much less constrained than the one I had been taking, making it actually doable.

I wasn't hot on ABC as a revealer, because 1.) ABC order is not as strong as ALPHABETICAL order and 2.) it came at the beginning of the grid, giving away the game early. In mine, I was planning on using THE ALPHABET SONG, which would come at a better location for a revealer. But like my overall approach, that also proved impossible to execute on.

Impressive that Alex finished the grid with not nearly as much crossword glue as I would have expected from this set of constraints. Yes, there did feel a lot when I solved — ESS, CRTS, HOI, HEE, COL, BBL, and more — but think about how difficult it is to fix so many letters into place, all throughout the grid. It's true that some letters could be shifted, i.e. from C to D or from R to S, while still maintaining alphabetical order. But a single move like that will affect the rest of the puzzle, forcing a cascade of changes.

And to work in some great bonuses! EVILDOER, I ADORE IT, LUCASARTS, BOOYAH! Well done there.

As a constructor, I was wowed — Alex's note about 76 words being easier than 78 words was particularly interesting to me. As a solver though, the effect was less impactful, especially for a Thursday puzzle, where I expect to work harder for a bigger a-ha moment.