So it's been about a week since half the internet lost its mind over a badly-lit photo of a dress from some wedding in Scotland, and I still haven't shut up about it to anyone unlucky enough to be near me. My wife is sick of it. My coworker Dave is sick of it. I don't care. I'm right and they're wrong and I'm going to explain why, because apparently that's what this blog is for now.
For anyone who's been off the grid: a photo of a dress showed up on Tumblr the Thursday before last, and people split into two camps almost instantly: some see it as white and gold, some (correctly) see it as blue and black. It's the same JPEG. Same pixels. Two people sitting on the same couch looking at the same phone screen will argue about it. I watched this happen in real time in a group chat around 11pm that night, and by the next morning it had eaten Twitter, Facebook, and apparently several newsrooms that should've had better things to do.
Here's the thing that actually got me interested past the initial "huh, weird" reaction: this isn't a trick photo and it isn't your monitor being miscalibrated (though, seriously, if you're reading this on a work laptop from 2011, get that thing profiled — mine was throwing everything a sickly yellow for months before I noticed). The dress thing is your brain trying to guess the lighting the photo was taken in and then correcting the colors based on that guess, before you're even consciously seeing anything. If your visual system assumes the photo was taken in shadow, it discounts the blue cast and you end up perceiving white and gold. If it assumes bright light, you keep the blue. Same raw data, different assumption, different answer. Nobody's lying about what they see. That's the unsettling part.
The color science writers had a field day walking through opponent-process theory and how the eye adapts to ambient light, and I'm not going to pretend I fully absorbed all of it on a Tuesday afternoon between meetings. But the gist stuck with me: the photo is genuinely, technically ambiguous in a way that most photos aren't, mostly because of how it's overexposed and how the light source isn't clear from the frame. It's basically an optical illusion that happened by accident instead of by design, which is a much better story than anything Photoshop related.
What I actually want to talk about, though, is the plumbing underneath the meme. BuzzFeed's post with the poll reportedly did enormous traffic that day, fast enough that it turned into its own little case study about serving a page under a real-time hit spike, the kind of thing engineers write up afterward with graphs. I love that half of this story is "existential crisis about the nature of perception" and the other half is "some poor ops person watching request-per-second dashboards climb at 1am wondering if the caching layer is going to hold." Both of those are real. Both of those happened because of a five-dollar dress from a shop in Scotland.
And can we talk for a second about how fast this thing got monetized? Within two days there were dress-themed t-shirts for sale. There's something very of-this-decade about a random photo going from "wait, what color is that" to a merch table in under 72 hours. I don't say that as a complaint exactly — more just, huh, that's where we are now. The internet doesn't let a moment breathe before it's a product.
I've gone back and looked at the photo probably thirty times since it happened, at different times of day, on different screens, trying to force myself to see white and gold just to understand the other camp. Couldn't do it. Not once. It's blue and black, it was always blue and black, the actual dress (which the store apparently confirmed) is blue and black, and I'm suspicious of anyone who tells me otherwise with total confidence. Which, if you know anything about how this argument works, is exactly the wrong attitude to have, because the whole point is that the other side isn't lying either.
Anyway. Back to actual work. Dave, if you're reading this, it's blue and black and you know it.