On Wordle and Fragmentation

So, the New York Times bought the word game phenomenon Wordle for ‘low seven figures,’ or expressed in more human terms, ‘upward of a million dollars.’ I’m happy that Josh Wardle got his bag, though I despise the NYT for things like rampant copaganda, warmongering, transphobic editorial practices, and puzzlingly enough, boot-licking anti-labor covid jokes. It seems logical that Wordle will eventually get wrapped in with the other games that they NYT bundles alongside its crossword section, itself mired in controversy. Needless to say, I’ll be opting out of the game as soon as opting in means the NYT gets any ad revenue or mined data. A cursory glance at the replies on the NYT’s awkwardly third-person-voiced tweet shows that I’m not the only one who is unhappy with the news.

Wordle’s viral success was pretty clearly due to a combination of only a handful of factors. It’s a quick game and it’s difficult to lose. There’s only one puzzle every day, and everyone in the world has the same one. Upon completion, a share card generates the ubiquitous and iconic emoji-based teaser. These things combine to make a game where everyone is on the same page, generally feels a sense of accomplishment, and can readily share a captivating glance of their progress with everyone else who is in the know. It’s an ideal model of a viral design, even if the creator just made it on a lark for his partner to enjoy. I don’t think this was crucial to its success, but it didn’t hurt that it was self-contained, a single HTML file calling a single JS1 file with the only unnecessary fluff being Google Tag Manager. Even if people didn’t realize this, it certainly had an impact as far as being reliably quick to load. Grossly, the NYT writing about it prior to buying it almost certainly helped as well.

In my mind, the acquisition news immediately and irreparably fragmented what felt like the most important aspect of the game – the fact that folks across the globe were all working on the same challenge over the course of the same 24 hours2. Debates over which clone(s) – of which there are many – we should all migrate to flooded the replies to NYT’s tweet, Discord servers I’m in, and text messages from my friends. Almost as quickly, conversations about the game seemed to dry up significantly. When there’s one clear canonical version of a thing that everyone enjoys, it’s easy to stay in sync. When enough people suddenly hate that version of the thing, we’re now plagued by the paradox of choice. Assuming we all collectively get over that hump, the best-case scenario is that those of us who fled Official Wordle all collectively agree on the One Canonical Clone. This won’t happen, of course; attempts at unifying on this cause are already comically fraught. But let’s pretend it does.

The New York Times is going to succeed with its acquisition; there are plenty of NYT subscribers, plenty of folks who subscribe solely to the crossword, and plenty of players who simply won’t care unless and until it disappears behind a paywall. In this best-case scenario, we’re still going to have two distinct groups. How this plays out depends on the hypothetical One Canonical Clone.

Wordle clones are a dime a dozen. As mentioned, the game has no inaccessible dependencies. It’s entirely contained in an HTML file and a JS file. You can go download these now and play a local copy forever3. As such, plenty of the clones out there are exact clones, or nearly-so, just changing a name or colorscheme. I know of several of these hiding in the shadows on personal servers, for obvious reasons. If we all agreed one of these was the One Canonical Clone, we’d maintain perfect parity with the Official Wordle, but taking it down would be trivial for the NYT and we’d all have to make the clone decision again. The (technically, not legally) open nature of the source also means that we have access to the two word lists that the game uses. It is thus entirely possible to write the game code from scratch, only scraping the word lists. It’s worth noting that the order of the solutions list matters as well – it’s prerandomized, and the script doesn’t rely on PRNG at all. This too would ensure parity between Official Wordle and the One Canonical Clone, but I imagine NYT could get this taken down via the word list copyright violation. They could also at any time obscure their own code and change either of the word lists and/or implement PRNG. Of course, plenty of the clones that folks are flocking to4 don’t even attempt parity from the jump.

This of course ignores matters of feature parity and deviations from the original. Does the clean-room One Canonical Clone have a colorblind mode? Can NYT sneakily claim grids of yellow, green, and grey are protected trade dress? Will the One Canonical Clone’s developer add “helpful” features that add just enough friction to make daily play more of a nuisance than a joy? The clones are all too similar to declare an obvious replacement, but a lot of questions would remain even if we all managed to unite on this.

I don’t know what the point of this post really is. I think sometimes we grow attached to a Thing and we lose it for reasons beyond our control. This sucks, and in a desperate attempt to cling to the Thing, we frantically scramble to replace the Thing while realize that what made the Thing special could only really exist within the Thing’s original context. Sometimes it’s best to learn from what you loved about the Thing and channel that into Something New. Enjoy Things while you have them, lament their loss, but know that you’ll experience that feeling again. In my opinion, the news of Wordle’s acquisition was enough to ruin the Thing. And I think that just has to be ok.


  1. Plenty of folks aptly noted that Wordle felt a lot like the good ol’ days of the internet. Functional, not bloated, not swarming with modals and ads. There’s a lot to be said for that. ↩︎
  2. Rollover is at midnight local time, so there is some disparity across time zones. ↩︎
  3. Okay, not technically. The solutions word list contains approximately 2,300 entries. These are not chosen with a PRNG; they’re hardcoded into the game in a random order. My understanding is that there isn’t a contingency once the end of this list is hit in 2028. ↩︎
  4. hello wordl is a prime example. Interestingly, this was a clone that existed prior to the acquisition, largely for the sake of being able to play more than one puzzle in a day. After the acquisition the dev changed it over to daily mode by default to more closely mimic the original game. I don’t mean to knock the dev here at all, she’s doing good work on a thing people want. The response to the news has just felt like a scramble from all sides. ↩︎