Under my thumbnail

In partnership with

Today’s advertiser is somehow, once again, Authory, whose automated portfolio system I’ve subscribed to for years now. Although Beehiiv rules prevent me from asking or encouraging you to click Authory’s ad, if you do so, of your own free will and according to your own moral principles, each click will earn me $2.

From the day I launched this project—November 1, 2024—until early February, I used AI imagery to illustrate it. I highly regret that decision, although I don’t know how I could have avoided it. The images, mostly produced by Dall•e, may at times have been fascinating, interesting, surprising, but they were never good. In fact, they were bad. I am glad to be rid of them.

For the past nearly two months, I have instead been mining public-domain art repositories for illustrations—and man, it is so cool! There are thousands upon thousands of high-quality, high-resolution images out there that are either in the public domain or available to use under a Creative Commons license, and if you are willing to think flexibly about how a piece of writing should be illustrated, there’s almost no end to the options.

Once I have an idea of my headline (or subject line, if we want to be precise), I take key words from it and go directly to the Metropolitan Museum of Art, the Art Institute of Chicago, the Getty, the Smithsonian, and the National Gallery of Art. (There are so many more!) All of them participate in a program called Open Access, which makes available to everybody high-quality records, including images, for works in the public domain. Of course, you get to filter by dozens of options, like format (I tend to opt for Paintings and Prints), “Available for download” (very important), and, if this is your thing, “Nazi-era provenance.”

The process has, so far, been incredibly rewarding for me, as I hope it has been for you. This really is a new way of discovering art that I connect with. When I’m in a museum, for instance, what I see is often out of my control—I wander from room to room, trying to appreciate the paintings and prints and etchings and sculptures that someone else, some curator, has decided belong in that particular spot. This isn’t terrible! I’ve encountered hundreds of works that speak to me that way, like this 1879 portrait of Joan of Arc by Jules Bastien-Lepage:

But going through the search mechanism is fundamentally different: I’m seeking an image that fits a particular theme, and because I can enter a basic term like “chicken” into a search engine, I have access to untold depths—and to aspects of my own aesthetic I did not know existed.

Apparently, I like Winslow Homer! I’ve used his works to illustrate two stories so far, I think because they have the right blend of specificity and evocation. This one may not be Key West, but it looks like it could very well be; whereas this one most certainly is of Santiago de Cuba, at a time of war and paranoia and looming darkness, which paired well with the essay. I’d heard the artist’s name before, of course, but never given him much thought. Now, though, he’s someone I pay attention to.

So, too, Goya, whose prints have adorned two essays so far. And Monet, a superstar so easily dismissed, worked to illustrate a poem. Because their works ostensibly suit something I’m already interested in, they come to me with meaning preordained. I understand them better, I think; I connect more easily. I’ve never been particularly artistic—I cannot draw—and so have often felt cut off from capital-A Art, but this brings me more in touch than I ever imagined.

More after the highly clickable ad…

Here’s how the process goes: I’ll plug in a search term. Today’s headline is “Under my thumbnail,” a Rolling Stones reference, so let’s see what “thumb” brings up.

Honestly, it’s not great.

Across all of these institutions, what we get is a lot of… Tom Thumb, the late 19th-century performer made famous by P.T. Barnum. Here’s one from the Getty:

Very cool image, but not necessarily one that frames this story properly.

The Smithsonian, meanwhile, has loads of scientific illustrations of birds (perched on scientists’ thumbs) and plant specimens I can’t really explain. The Getty has a great Robert Mapplethorpe (SFW!) and a Basquiat, but neither is public domain or Creative Commons, so I can’t legitimately use them. Nothing I’m finding is all that impressive or meaningful, I have to say, and I’m tempted to search instead for “nail,” whose initial results are quite surprising—but then, at the Getty, I happened on this:

It’s an 1885 painting of young Jeanne Kéfer, by the Belgian artist Fernand Khnopff, and it’s almost exactly just not right—but not quite. That thumb, that’s it right there. It makes the painting, elevating its subject from a bourgeois little girl to a child with possibly Napoleonic tendencies. It sticks out, as thumbs are wont to do.

But in the context of a newsletter and web post, it’s far too subtle. Though you’d be cued from the headline/subject line to think “thumb,” it’s easily lost in the composition, especially for a quick-glancing, speed-reading web recipient like you, my dear subscriber. Luckily, though, the Getty provides the image at a ludicrously high resolution—7434×7433, which is a delicious measurement in itself!—so I realize I can crop it into relevance. That’s what you saw up at the very top of this essay: a detail of that painting that was precisely 1200 pixels by 630 pixels.

Why 1200×630?

This is a very good question—and one I have not been fully able to answer. But here goes!

If you have worked in social media over the last decade or so, you will instantly recognize the specs: 1200×630 is the recommended image size for posts on Facebook, and on many other social media platforms. The resolution is high enough to provide a good level of detail within the image, no matter if it shows up on a desktop monitor or on someone’s phone. It will not turn fuzzy or pixelated unless you zoom way in, which you probably won’t. 1200×630 seems reasonable.

It’s also completely nonsensical. Photo and video formats are defined by their aspect ratios—the ratio of their width to their height—for purposes of standardization. Up to the early 1950s, photographs and motion pictures generally had an aspect ratio of 4:3, or slightly wider than they were tall. This was fine. But in the 1950s, when Hollywood realized it had to compete with the new technology of television, studios debuted new formats such as Cinemascope, an ultra-widescreen 2.35:1 projection that required special lenses to display properly. Think Lawrence of Arabia or Star Wars. A less-extreme format at a 1.85:1 aspect ratio was introduced around the same time, and since it didn’t require as much special tech to show in theaters, it eventually won out.

Most movies produced since the 1970s are 1.85:1. We’re used to it, and it makes sense. Our eyeballs—or at least mine, IDK about yours—sit side-by-side, and 1.85:1 suits that biological geometry pretty well. If you’re used to it, watching a pre-1970 movie or a pre-HD television show can make you feel a little cross-eyed. Those frames look squished, claustrophobic, but at least 4:3 makes numerical sense.

However: 1200×630 is an aspect ratio of 1.90:1. Why? WHY?!?

As far as I can determine, 1200×630 was chosen as the proper image size back in 2010, when Facebook first created the Open Graph protocol. Open Graph is actually really smart. Back then, everyone wanted to share with their friends cool sites they found on the Web. But how was a system like Facebook—and every other social media platform, plus lots of other web systems—supposed to understand what was on the pages that everyone was sharing? The technology wasn’t sophisticated enough to, like, read the pages and make sense of everything on them (and it still isn’t), so Facebook suggested a standard approach for every website to use to let everyone know, in the HTML code of every page, the headline, the description, the thumbnail image, and so on. This was Open Graph, and here’s what it looks like for Jeffrey Goldberg’s recent Atlantic scoop on the national-security Signal chat group:

<meta property="og:title" content="The Trump Administration Accidentally Texted Me Its War Plans"/>
<meta property="og:description" content="U.S. national-security leaders included me in a group chat about upcoming military strikes in Yemen. I didn’t think it could be real. Then the bombs started falling."/>
<meta property="og:url" content="https://www.theatlantic.com/politics/archive/2025/03/trump-administration-accidentally-texted-me-its-war-plans/682151/"/>
<meta property="og:type" content="article"/>
<meta property="og:image" content="https://cdn.theatlantic.com/thumbor/3XBevuI50bX2JBa7sezKu1AaFCc=/0x440:5990x3560/1200x625/media/img/mt/2025/03/GettyImages_2204323444-1/original.jpg"/>

Each of those lines spells out a different aspect of the story: the headline (or title), the dek (or description), and the URLs for the article itself and the thumbnail image that represents it. Importantly, these don’t have to be the same as the headline, dek, and image visible on the article, which is why sometimes you’ll see a story framed one way on Facebook or LinkedIn and another once you’ve clicked through. The editors have crafted display copy and images that they know will create a reaction on a social media platform, but once you’ve arrived on the news site, a different approach is required. This is not deception—this is good journalism and audience development. Ask me about it sometime!

None of it, however, explains why the Open Graph committee, which was created by Facebook, chose 1200×630 as its baseline image size—an aspect ratio in use nowhere else in media! They could have chosen instead 1200×648, which would have preserved the popular 1.85:1 format, or even 1200×650, which is almost indistinguishable, but instead they went with 1200×630, which is just barely off enough to require intelligent cropping of 1.85:1 publicity stills from movies.

The problem with 1200×630 is not its irregularity in the context of the last 50 years of film and TV. It’s that that narrow horizontal slice is out of step with thousands of years of art that preceded it. For much of human history, paintings were as often vertical² as they were horizontal—the former for portraits, the latter for landscapes¹ . But now, in a social-media context where thin horizontals rule³ , we’re required to produce crops that merely allude to the original artwork. In some cases, like today’s thumb-themed image, that lets people like me home in on a significant detail. But it also leads to auto-crops like this:

Which I suppose in this case is appropriate! But still not ideal.

There are things I/we can do. I could have produced multiple versions of that “bad hair” image to suit the different ways the story might be presented. Or Beehiiv could implement a “focal point” system where I could designate a point on the image which the auto-cropping system would attempt to highlight. Either way, though, we’d continue to labor under the tyranny of the current system, whose arbitrary ratios have no explanation, no justification, beyond the fact that everybody now accepts them as the price of putting content on the Internet.

For a couple of weeks now, I’ve been trying to get answers from the source. I’ve been emailing Facebook’s media contact inbox—[email protected]—asking to speak with anyone, anyone, who was involved in the original Open Graph talks. I’m not angry, honest I’m not; I’m just curious. But so far I’ve got no response at all. If you, too, are curious, or if you just want to help put my curiosity to rest, please email them yourself to demand that they speak with me. Trying is the least we can do.

Although, to be fair, once I get Meta on the line, I may have one or two more questions for them. 🪨🪨🪨

Read a Previous Attempt

1  Though rarely, in my experience, as widescreen as 1.85:1.

2  I don’t have any actual statistics to back this up, but maybe you do? Someone does?

3  Except, of course, on Pinterest, TikTok, and Instagram, where verticality rules—but where thumbnails are less defined.