Category: Winning Work

The first video was posted by a girl called @kaylalikesitloud and it was about a milkshake. Twenty-two seconds. She holds the cup up to the boardwalk lights and rotates it so the camera can admire the strata—cookie crumble, soft-serve, a wedge of funnel cake planted on top like a flag. Behind her left shoulder, out of focus, my brother Danny is leaning on the railing, looking at the ocean.

The video was filmed June 9. He died June 30. The algorithm showed it to me on July 21, between a man restoring a hand plane and a woman who wanted me to quit sugar.

I watched it forty times that first night. Not the milkshake. The eleven frames where his face almost turns toward the lens. He’s wearing the green shirt I put in a contractor bag in August with the rest of his shirts. He is doing nothing. That was the unbearable part. He is just standing there being alive, and at his back a girl is filming a milkshake, and neither of them knows anything.

I learned the grammar of the search. The boardwalk has a geotag; the geotag has a page; the page had, that summer, 3,400 videos. Ring toss. Sunburn. A hermit crab named Sergeant. A proposal, two fistfights, a hundred more milkshakes. I watched all of them. I watched them the way you’d pan a river, and six times the pan came up gold.

June 9, the milkshake. June 11, he’s a smear crossing a skateboard video by the arcade; June 14, he’s in line behind a woman reviewing a taco stand, reading the menu board with his weight rolled to the outsides of his feet, the way he stood in every line of his life, the way all his sneakers wore out at the edges first, and when she pans to the salsa bar he’s gone. June 17, he sits alone on a bench at the edge of a family’s beach day, forty feet of sand between him and the camera, and for the length of the video he doesn’t move. 

June 22, he’s laughing. This is the one I can’t watch anymore. Someone’s dashcam-style walking video, the kind with no subject at all, just a man narrating his steps for reasons the internet no longer requires him to explain, and Danny passes him going the other way, phone to his ear, laughing at something the phone said. The audio is a sea shanty. 

Every video, the audio is something. A shanty, a sped-up pop song, a voice explaining the milkshake. Six videos of my brother’s mouth moving and not one second of his voice. I downloaded them all and muted them and played them in a row, and it was like watching him through the window of a train.

The sixth is June 28. He’s at the far end of the frame of a girl filming the sunset, and he is also filming the sunset, holding his phone up to the same sky, and I have been through his account, his drafts, his camera roll, and there is no sunset. He never posted it. He never even kept it. Whatever he saw that made him raise the phone, he looked at it and deleted it two days before, and I want to ask him what was wrong with it. That’s the question, of all the questions. What was wrong with the sunset, Danny.

By September the algorithm understood me completely. It’s not a mystery how. I had watched twenty-two seconds of milkshake 400 times; I had lingered; the machine measures lingering. My feed became boardwalk, then Jersey shore, then strangers’ background figures everywhere, as if it had grasped not the words of my search but the shape of it—that I was looking for a man at the edges of things. It started offering me other people’s Dannys. Out-of-focus fathers. Men on benches. 

Once, a woman’s video about her divorce, and behind her in the coffee shop a man with his head in his hands, and I watched it eleven times before I understood he was a stranger, and I watched it twice more after that.

Then October 4. Posted that morning by an account that reviews diners. A booth by the window, pancakes arriving, the reviewer saying okay so first impressions. In the corner booth behind him, half out of frame: the green shirt. And in the aisle, one sneaker, tipped onto its outside edge. The diner is in a town Danny never went to, and the video is dated ninety-six days after we buried him.

I sent the account a message. I kept it casual, the way you’d keep your voice casual asking a nurse a question you already know the answer to. Hey—random, but do you know the guy in the corner booth behind you? The reply came at 1:40 in the morning, which is when everyone is awake now.

what guy lol

the signal is best when the rest of the street
is already sleeping, 

when mothers have finished shouting 
their last prayers into the rafters. 

i sit on the porcelain edge of the bidet, 
the screen glow making a small blue chapel 

out of the bathroom tiles. 
the app asks for my location. 

i tell it abuja, nigeria.

i tell it two kilometers away.

my thumb drags the grid upward,
once, & then again.

because the night opened its wide, 
empty hands

& filled them with nothing but hours, 
i jump into a waterless well.

click—
another torso,

click—
another soul,

click—
floating past my screen.

a profile appears with no face,
only an image of a black car wheel & a name 

that is just an initial. 
under interests, it says clean & masc pls.

he sends a message: are you discreet?
which is the modern way of asking: 

are you ready to die a little bit 
without making

a sound?

i look at my feet on the cold terrazzo floor. 
my toes are the exact shape 

of my grandfather’s. 
a long line of men who crossed rivers on foot, 

who killed leopards with flintlock guns.

& here i am, at the end of this bloodline, 
trembling in the dark 

because a stranger across the highway
asked if i know how to lock a door

& keep a secret.

the same god who watches the leopard fall
watches me type yes.

“I’m making traditional fajitas tonight. Can you give me an authentic recipe?” Erin doesn’t ask the chatbot because she wants a recipe. She asks because she wants evidence.

Three little dots pulse.

Then:

Certainly! Traditional fajitas are widely believed to have originated in South Texas among Mexican and Mexican American ranch workers. A simple preparation includes skirt steak, lime juice, garlic, black pepper, and salt…

She smiles.

There it is.

Salt.

She takes the screenshot before she finishes reading the paragraph.

Then another with the timestamp visible.

Just in case.

***

She posts.

Asked AI for an authentic indigenous Mexican recipe.

First ingredient after beef?  

Salt.  

Colonialism literally baked into the training data.  

Within four minutes someone reposts the screenshot with SALT circled in red. Someone else adds a crying emoji over the steak.  

Within eight minutes another version appears. Everything after salt is gone.  An account called DecolonizeYourDinner replies:  

AI perpetuates systems of oppression because it is trained on archives produced by empire.  

Erin likes this.  

Someone else writes:  

Exactly. Thank you for documenting this.  

She likes that, too.  

She tells herself she’ll stop when she reaches the bottom of the replies. When she reaches the bottom, there are more.  

A blue verification check appears next to one of them.  

There is no such thing as a neutral recipe.  

Erin has no idea who the verified person is. She likes the post.  

***

Soon the replies are just replies to other replies.

The first long thread explains that salt was historically connected to extraction. 

A second argues that isn’t even what extraction means.  

A third insists correcting terminology is inherently extractive.  

By evening, a forty-post thread questions whether authenticity is itself a colonial framework.  

Nobody mentions onions.  

***

Someone whose profile picture is just his forearms flipping something on a grill replies. 

Fajitas are Tex-Mex. Salt is a completely normal ingredient.  

No likes.  

Someone quotes him.  

Salt is normal now? Wild.  

Another asks him to examine why he feels so comfortable centering whiteness in culinary conversations.  

He replies:  

I’m Mexican.  

For almost a minute, nothing.  

Then:  

Internalized colonialism is still colonialism.  

Erin reposts that one.  

***

Erin refreshes.  

Then again.  

Before the new replies finish loading, there are newer ones.  

The screenshot becomes an infographic tracing each ingredient through histories of colonial dispossession.  

Then a slideshow ending with: Food remembers. Algorithms don’t.

The first person to say palate laundering puts it in quotation marks. 

The second asks if it should be hyphenated.  

The next four don’t use quotes.  

The chatbot has allegedly erased thousands of years of culinary history, committed digital land theft, and engaged in algorithmic palate laundering.  

Erin uses the term in a sentence.  

Nobody asks what it means.  

***

Late that evening a food historian joins the conversation.  

She has spent twenty years studying the foodways of Indigenous communities and early ranching cultures across the borderlands of northern Mexico and the American Southwest.

Her thread is patient.  

Almost irritatingly patient.  

She explains that beef itself arrived with Europeans, but cattle culture evolved locally over centuries.  

She explains that early fajitas emerged among vaqueros.  

She posts photographs of handwritten recipes.  

Every one contains salt.  

There are citations.  

Scans from archives.  

Primary sources.  

Even a link to the university press book she’d written on the subject. 

Eleven likes.  

One reply.  

Source?  

The historian links to her book. Again.  

The reply is deleted.  

***

Erin’s original post passes 100,000 likes overnight.  

She wakes to interview requests.  

A podcast.  

A public radio producer.

Two journalists.

One asks whether she believes AI has become a vector for culinary colonialism. 

Erin isn’t entirely sure what that means.  

“It definitely raises important questions,” she says.  

Thirty minutes later, it is a headline.  

Erin updates her bio.  

***

Someone asks the chatbot the same question.  

The recipe is unchanged.  

Beneath the ingredients, it now adds a footnote:  

Food traditions reflect diverse cultural histories and lived experiences. No single recipe can fully represent the communities and traditions associated with this dish.

The footnote gets its own screenshot.  

It spreads faster than the original.  

***

Three days later there are shirts.  

NO ALGORITHMS IN MY KITCHEN.  

Erin orders one.  

A Michelin-starred chef resigns after screenshots of his chatbot conversations are leaked.  

An international coalition of food justice organizations forms. Its mission statement promises to honor culinary traditions as living records of community memory, shared  stewardship, and collective accountability.  

In a televised hearing, a congresswoman calls AI-generated recipes “a break in the chain of our shared cultural inheritance.”  

Erin’s interview is entered into the Congressional Record.  

Nobody ever tries the recipe.