Kitchen weather

You are in the produce aisle, on an ordinary Tuesday. You came in for three things and you will leave with nine. The phone in your pocket says rain tonight. But the aisle is a weather report too, if you know how to read it. What is piled high and cheap this week. What is small, pale, and priced like jewelry. What is missing entirely. Strawberries in February. A cauliflower that costs as much as a chicken.

Most of us read this weather without knowing we read it. You do it with your peripheral vision. The avocado pyramid means guacamole is on this week. The picked-over shelf means the sale was good. The gap where the local greens should be means the truck is late, or the season is short, or something happened upstream that you will never hear about. You adjust your dinner before you notice you have adjusted. That instinct, the shopper's read of price and pile and gap, is the oldest data science in the food system. Cooks have done it forever. What has changed is on the other side of the shelf: the reading is now done industrially, by the store itself, with instruments that see every basket in every store at once.

Raj Thandhi reads this weather for a living. She is a chef, a recipe developer, and a food educator, and a large audience cooks along with her. On the Future Herd podcast she told us about the day the weather beat her. She had posted a recipe she believed in, and it bombed. Nothing was wrong with the recipe. The problem was the cauliflower: it had gotten so expensive that week that most of her audience simply could not go there. "I didn't read the room," she said. "It's not the right time to be asking people to buy that vegetable."

That is a strange and wonderful thing: a recipe is a small forecast. It says: here is what you could cook this week, with what the season and the store are offering. Raj's forecast failed because the price moved faster than her read on it. She now describes her work as a constant feedback loop: watch the prices, watch the season, watch the mood, adjust. She reads the kitchen weather every single day, by feel.

She is not the only one reading it. The store is reading it too. The difference is that the store reads it weeks earlier, with machines, faster and wider than any person can. That difference (when the reading happens, and who it serves) is where this chapter is going.

The store is a forecast

Everything on that shelf was predicted before you arrived.

Not guessed. Predicted, in the formal sense, by software whose whole job is to know what you will want before you do. Consider Loblaw, the largest grocery retailer in the country. For more than ten years the company ran its own demand forecasting, the quiet math that decides how many cases of what go to which store. Ten years in, they hit a ceiling: the forecasts were as good as the old methods could make them. So they brought in machine learning, software that learns patterns from examples instead of following instructions written in advance, in the form of a platform called Luminate Demand Edge, built by a supply chain software company called Blue Yonder. The machine learning broke through the ceiling the old way had hit.

Translate that out of the business page and into the aisle. The model reads millions of past baskets: what sells together, what sells before a long weekend, what sells when the price drops ten cents, what sells when it rains. Then it tells the store what to order, how much, where to put it, and when to mark it down. The pile of corn in July, the wall of soup in October, the exact week the barbecue supplies appear: these are not instincts. They are outputs. The shelf tonight is the physical form of instructions written last week.

You can catch the forecast reading the sky if you watch the aisle across a season. The barbecue supplies go up before the first warm weekend, not after it. Soup weather arrives in the store before soup weather arrives outside. Ice and lemonade multiply ahead of a heat wave the way sandbags multiply ahead of a flood. The store reads the weather as carefully as any farmer does, because you change with the weather: your appetite, your plans, your idea of a reasonable dinner. The store's whole business is knowing your Tuesday better than you do, and the sky is one of its best clues. Long before we get to a single sensor on a single field, the food system is already a machine for turning weather into behaviour.

And that, in plain language, is what this book means by programmable, before a single drone or sensor or robot has entered the story. A programmable thing is one you can keep giving instructions to, long after it is built. The grocery store used to be a building with habits. Now it is a building with settings. It looks the same from the parking lot. The instructions changed.

It is easy to slide past, because we tend to picture technology as objects: machines, screens, robots. But the most important technology in your grocery store is invisible. It is the forecast, and the forecast is not a thing you can point a camera at. It is a relationship between the past and the future, written down in code, running every night, rearranging the shelves you will walk past tomorrow. You have lived your whole adult life inside such forecasts. This book is simply going to follow them to their source.

Their source, meanwhile, is multiplying. Not long ago, agriculture's data problem was scarcity. Mohamad Yaghi, who leads the Innovation Hub and AgExpert at Farm Credit Canada, dates the turn precisely: "As opposed to 2019 where there wasn't really a lot of data to collect, we're now in a scenario where there's too much." On the Future Herd podcast, Mohsen Yoosefzadeh Najafabadi, a University of Guelph professor who works on AI in plant breeding, put the new problem plainly: we are all collecting everything now; the question is how to "get the best juice out of whatever we are producing in terms of the data."

The space between those two sentences is where this book lives. The store already squeezes its juice every night: that is what the demand forecast is. On the farm, the juice mostly sits unsqueezed, in machines and apps and notebooks and heads. Whoever learns to squeeze it gets to help write what happens next, on the farm the same as in the aisle.

The forecast gets a face

Lately the forecast has started talking back.

In February 2026, Loblaw put its grocery app inside ChatGPT: you can now plan dinners in conversation, describe what you are in the mood for, and have the conversation produce a basket from your local store. The company presents it as the first grocery shopping app built directly into a chatbot. Behind the scenes, store owners and managers have their own assistant, called Robin. In the United States, Kroger has announced a shopping assistant built on Google's AI, and every other large grocer is making a similar bet with a similar partner.

The details of who partnered with whom matter less than the direction. The grocery list is becoming a conversation, and every conversation has two ends.

Here is the question riding in the cart: who is the forecast for? The model that suggests your meals knows your basket, your budget, your Tuesday night. It can make your week easier; that part is real. It is also, and first, an instrument of the store. It was built to sell groceries, tuned to sell groceries, and judged on whether it sells groceries. None of that makes it wicked. It makes it legible. The scholar Donna Haraway has a way of putting this that I keep coming back to: all seeing is from somewhere. Every way of looking has a home address, and the ones to watch are the ones that present themselves as the view from nowhere: pure math, neutral, inevitable. A demand forecast is never the view from nowhere. It sees from the store's somewhere, and it is tuned to the store's purposes. Predictions are always made for someone, and it is worth forming the habit of asking for whom.

There is a second habit worth forming, and it is noticing who the forecast assumes you are. The new conversational grocery tools are built for a particular shopper: comfortable with a smartphone, fluent in the app's language, well resourced, predictable. Nobody is building the conversational store for the person who shops by feel and by cash, or for the newcomer whose cooking does not match the model's picture of this country. Raj's audience lives in that gap. When a whole cuisine runs on vegetables the forecast considers niche, the forecast quietly decides whose dinner is normal. All of this lands, remember, in the middle of a loud public argument about food prices, an argument the big grocers are actively trying to manage. The technology is not separate from that argument. It is the argument, continuing by other means.

This book picks that thread up again when we reach the fork in earnest.

The real weather

Now the turn that gives this chapter its name.

You checked the weather this morning. So did every farmer you have ever bought food from, and they checked it harder, earlier, and with more riding on the answer. The app on your phone and the decisions on the farm are fed by the same instrument: a planetary machine of satellites, sensors, stations, and models that reads the atmosphere and writes the future a few days at a time. More and more of that machine's sensors now live on farms themselves, bolted to barns and fence posts, sipping from the same data streams your phone does.

This instrument is one of the quiet great accomplishments of public life. Most of its foundations (the satellites, the ocean buoys, the weather stations, the models that hold it all together) were built over decades with public money, by public agencies, for everyone. The forecast on your phone is a commons you carry in your pocket, so ordinary that we forget it is extraordinary. Farmers have never forgotten. Their work runs on it. And in the chapters ahead, as the farm itself fills up with instruments, the comparison holds: the parts of this system that were built for everyone behave very differently from the parts that were built for someone.

The forecast on your phone tells you whether to carry an umbrella. On the farm, the same forecast decides whether to plant, spray, irrigate, harvest, or wait. One system, two ends of the food chain. The weather app is the only piece of agricultural infrastructure most of us will ever touch, and we touch it every day, usually to decide what to wear.

There is a complication, and you have felt it without needing a graph. The weather has stopped repeating itself the way it used to. The old averages, the almanac wisdom, the grandparent certainties: they lie more often now. A later chapter will spend real time in that storm. For now, what matters is what it does to the two weather reports in this chapter. When the sky gets stranger, the forecasts get more valuable and less certain at the same time, for the farmer and for the store alike. Kitchen weather and field weather turn out to be the same system seen from two ends of one chain. Raj's bombed recipe and a farmer's replanted field are the same event at different distances: a forecast meeting a world that has changed its mind.

In the same conversation, Raj said something that stays with me. Describing what cooking should be about, she put it this way: "engaging with your food, being kind to the community that's building your food." That is not optimization. It is relationship. She reads the kitchen weather in order to stay in touch with where food comes from and who it comes from. The store reads the same weather in order to move cases. Same signal, opposite postures, and the difference between them is the difference between reading the forecast and being read by it.

Which brings the chapter back to its small, strange beginning. A recipe is a forecast. A shelf is a forecast. A farm is a forecast with a mortgage. The whole food system, from the soil to your Tuesday, is a stack of predictions about the future, each one written by someone, each one serving someone. Learning to read that stack, and eventually to write in it, is most of what this book is about.

The Brazilian educator Paulo Freire had a name for this kind of reading, and he built his life's work on it with people the official system had decided were not readers at all. Literacy, for Freire, was never only reading the word. It was reading the world: learning to see the forces standing behind the facts of your day, because once you can read them, you can start to write. That is the literacy this book is after, for the food system: the aisle, the forecast, the weather, and then the writing back.

Open questions

When the shelf is a forecast, where does the unplanned thing come from: the new ingredient, the dish a culture brings with it, the thing nobody buys yet but somebody would love?

Does being well predicted make us easier to feed, or easier to steer? How would we know the difference from inside the prediction?

If the same weather instrument runs your phone and the farm's decisions, whose instrument is it? Who maintains it, who pays for it, and who gets to ask it questions?

What would a forecast look like if it were built for eaters instead of for stores?

Agency moves

Catch one prediction. Sometime this week you will be offered a substitution, a coupon, a "you might also like," a basket built from your last one. Stop when it happens. Ask what the offer knows about you, and who it is working for. You will not get a full answer. Getting in the habit of asking is the move.

Read the weather yourself. Once this week, buy the thing that is piled high and cheap, and cook that. The pile is the system working in your favour for once: abundance is information, and the floor is broadcasting it. Raj reads that signal every day. So can you.

Ask one question the model cannot. Next time you are in the store, find a person and ask about one item: where it came from, when it arrived, why the other thing is missing. Watch where the knowledge lives, and notice where it has already left the building.

Situation: One ingredient, all the way home

Pick one ingredient from your kitchen this week. Something ordinary: a potato, an egg, a bag of rice. A cauliflower, if you are feeling brave.

Trace it home. Start where you found it and walk it backwards in your mind and, where you can, in fact: the shelf, the stockroom, the truck, the warehouse, the distributor, the packer, the field. You will not get all the way. Nobody gets all the way. That is part of the game.

At each step, ask two questions. What is the first computer in this part of the story: the scanner, the ordering system, the demand forecast, the route planner, the irrigation schedule? And what is the last human: who actually touched this, chose it, judged it, decided it was good enough to move along?

Then mark the spot where your trail goes dark. The step you cannot see into. The company whose name you cannot find. The question nobody in the building can answer. That dark spot is not an accident and it is not a failure of your detective work. It is a design choice, made by someone, about what you get to know. Later chapters are going to spend a lot of time in those dark spots. For now, just notice that you can find the edge of your own knowing in less than an hour, with one potato.

This one plays well with company. Same ingredient, different kitchens, different stores: compare trails over coffee or in a group, and compare especially where they go dark. Yours will go dark somewhere different from your neighbour's, and the difference between the two maps is the beginning of reading the whole system. Which is to say, the beginning of this book's real subject.

The aisle is a weather report. The store is a forecast. The forecast is written by someone. All three travel with us, because next we follow the weather to the field, and the instruments get wings.