This one began as an abstraction in need of a theme. My original, vague notion was using [individual blocks as stand-ins for letters] in some way. From there it was a matter of working backwards towards a fitting revealer. Going off a list of BLOCK phrases, I was lucky enough to find one receptive to playful reinterpretation.

Because it was the only bit of theme material, the 15-letter revealer had to go across the middle. For the sake of lucidity, it also felt best if the four ER blocks were totally isolated in all eight directions. These two constraints imposed a rigid block pattern on the grid, which, in its own way, was strangely comforting.

Oftentimes, when there are a myriad of options, it's easy (for me, at least) to get bogged down and expend way too much effort desultorily, without a set plan or course of action. Indecision, thy name is Tim. But when you're working within an inflexible structure, with an embedded central entry crossing multiple themers, you're forced to overcome problems, rather than discarding grid/fill #2836 and starting over again, convinced that the perfect setup is just over the next hill. It's the obsessive, insomniac crossword fan's equivalent of a straitjacket and a padded cell. Restraints which, judging from the comments, some solvers wish I were permanently kept in. Someday!

Can't remember a tougher Thursday! I was still hitting my head at the 30 minute mark. Such a relief to figure out that STUMBLING BLOCKS were interpreted as ER hidden inside four black squares. I've fixed up the database answers and added ERs (in white) below to make things more clear.

Check out all the interlock. I'm usually not one to be impressed at themer crossings, but there's so much of it I could barely get my head wrapped around it. I highlighted all the theme answers (in blue) below to give a better picture of it. Granted, Tim did have a lot of flexibility because there are many phrases/words starting with ER, but wow. Audacious construction.

I would have liked all the short themers to be more solid, though. NETH(ER), DIN(ER), CHATT(ER), (ER)ROL, (ER)ASURE are all fine. (ER)NES, SHO(ER), even (ER)IE PA felt weaker. Tough to do, especially since there aren't nearly as many words starting with ER as there are ending with ER. Perhaps a little less interlock would have facilitated better short stuff like (ER)IN, (ER)OS, (ER)DOS, etc.

As if his task wasn't hard enough, the grid layout — featuring shorter answers — meant that he had to work in a lot of long fill. AXIS OF EVIL is colorful, and HYGIENIC is pretty nice. ULTRAHOT feels slightly made-up, like ULTRA NICE or ULTRA SMART or something.

ANNE SEXTON doesn't get as many Google hits as I'd like to see (500K), but I'd argue that winning a Pulitzer makes you crossworthy. I'm not wild about that final N though, crossing ASANA. ASANA is a yoga term I see all around Seattle (we're a yoga town, but I think that crossing sets up too many solvers for a pure guess. For a puzzle that's already ultra hard (ha) to solve, I'd want to make more sure that people who stick with it are properly rewarded.

It was tough for me to remember which black squares were the ER ones, so it would have been nice to have them distinguished somehow. How cool would it have been if those four squares were somehow made to look like they were stumbling! Not sure how to do that, but fun to think about.