Indie crossword constructor here with what is now my fifth puzzle for the New York Times. Somewhat surprisingly, it's the first one with normal symmetry; it's also the first one that's a freestyle, rather than a themed puzzle. And I'll be honest, it's not my favorite of the bunch. Nothing wrong with that; not every puzzle can be the best. And this doesn't mean that it's a bad puzzle either; I certainly don't think so. But I'm always striving to make my puzzles better, and I do think there's some things I could have done better when I made this (and would do better if I was remaking it today).
That said, I really like the center here, which (unsurprisingly) is where this puzzle started. This puzzle came from a period of playing with common layouts, and I liked how the two stagger stacks came out here. Sure, some of the stuff is a little iffy: TEENTSY, I MUST, and ANI with a cluing angle that (as both a solver and a constructor) I hate, and doubly so because it could (and, IMO, should) have been clued as a name. But the long answers are great; at the time of construction, all but MEAN STREETS would have been NYT debuts (though I got scooped on FASHIONISTA), and all but RESTRICTIVE are pretty interesting and add a lot of flavor to the puzzle.
But RESTRICTIVE kinda sums up what this center did to the rest of the puzzle. It was rather hard to get the ring of sevens (BAT NEXT / BEER HAT / TEENTSY / TACITLY) to work, since all four of those entries have to cross three elevens and still work with two of the other sevens. This, in turn, constrains the five letter answers: four of them (SPARE / I MUST / BEGAT / CHRIS) must cross a stagger stack while also working with the sevens that are right next to them, and then the other eight five letter entries now have two letters locked in place, with four of them (LOTSA / EBERT / JUNTA / LSATS) still needing to work with the ends of the stagger stacks so as to feed into the corners.
The point being, each of these constraints further constrains the next entries out, and that's how things like LSATS get forced into the puzzle. And having that then constrains the corners; EV?????? has a lot more possibilities than EVA?????, which allows for more flexibility and freedom and makes it much more likely that you can have sparkling fill / interesting entries in the corners.
Still, for the most part, it wasn't too constraining, but I definitely feel like there was a lot of unmet potential for those corners. Even with interesting things like MATHLETE, SKYNET, BANK ON IT, FUN HOME, ARIANA Grande, and LIL KIM, there's things like ATS, RETAGS, SONARS, and I PUT; is it worth it? Maybe; some solvers won't mind, some solvers will. But as a constructor who takes pride in prioritizing clean fill, I definitely think I could have done better.
In any case: I'm not trying to beat myself up here; like I said, I don't think this is a bad puzzle. Mostly, I'm trying to explain to solvers (and especially non-constructors) how this was made, and offer a look behind the curtain at some of the hidden challenges of construction that aren't always apparent in the finished puzzle. But I also want to make it clear that constructors (myself included) should have high standards, and that they should always strive to do better, even if the puzzle's already good enough.