Eyes in the sky

In a house in Toronto, in the morning, a man opens his phone and walks his farm.

The farm is not outside. It is not in Ontario, or anywhere near it. It is his family's olive and grape farm in Lebanon, where his morning is already their afternoon, and he walks it the way you or I check the weather: aerial views from drones that have flown the rows, readings from sensors standing in the soil, a quiet feed of the day on the other side of the world. With AI-powered drones and smart monitoring, he has said, he gets real-time aerial views of the place. From his kitchen, he keeps an eye on his family's trees.

The man is Mohamad Yaghi, and we have met him before, in the first chapter, warning us that agriculture's data problem has flipped from scarcity to glut. By day, he leads the Innovation Hub and AgExpert at Farm Credit Canada, building data tools for Canadian farmers. In the off hours, he farms at a distance. He told the story on The Extensionists podcast in March 2026: how the technology makes it possible, and what farmers on this side of the ocean might learn from it.

The picture is easy to slide past. The eye is here. The farm is there. Twenty years ago that sentence was magic. This chapter is about how ordinary it has become, and about the question that rides inside it, the same question that followed us out of the grocery store: whose eye is it?

From novelty to infrastructure

Not long ago, a drone over a farm was a toy, or a gimmick at an agricultural show. Then the jobs started piling up.

Scouting: find the stressed patch of field before the stress spreads. Mapping: the whole farm rendered in more detail than any fence-line memory can hold. Spraying: not the whole field, just this row, just this weed. Searching: the missing cow, found by thermal camera at dusk. Assessing: after the storm, where the water went and what it took with it. None of these jobs is science fiction. They are Tuesday.

Strip the romance off the machine and a drone is a flying sensor platform. The flying is the easy part, and it gets easier every year. The real machinery is the pipeline behind it: collect the images, move them, process them, turn them into advice a person can use before lunch. When this chapter talks about drones, that pipeline is what it means. The wings are only the delivery system for the seeing.

And the eye keeps getting sharper. There is a company called Taranis whose specialty is leaf-level photography: from a drone, a single leaf, close enough to read what is starting to go wrong with it. One flight collects something like ten thousand photographs. The company's models learned to see by studying tens of millions of images, and behind that learning sits a detail I love, because it gives the game away: a team of thirty agronomists whose job is to tag pictures, teaching the software what disease looks like, one leaf at a time. The artificial intelligence turns out to have a staff room.

The company's pitch is that up to forty percent of crops are routinely lost to insects, disease, weeds, and nutrient problems, and that the eye lets you intervene earlier and smaller. That is a vendor's number, and you should hear it as one. But the direction is real, and every agronomist will confirm it: the earlier you see trouble, the cheaper it is.

Where the eyes live

So how many eyes are up there? There is no honest answer. The big numbers come from the biggest makers and their own reports, and nobody independent is checking them. So this book will not count the eyes. The better question is where the seeing lives, and that question has a concrete answer worth knowing.

Most professional drone work, the data it produces and the models that read it, is concentrated in wealthy countries and large companies. There is a network built to move that capability instead of selling more machines into its shadow. WeRobotics is a nonprofit, and its project is the Flying Labs Network: independent, locally led hubs in close to forty countries across Africa, Latin America, the Caribbean, and Asia-Pacific. A Flying Lab is not a branch office. It is a local business, self-financed, selling drone, data, and AI services to communities, governments, researchers, and nonprofits near home. Agriculture is one of its named trades. The Tanzania hub runs drones-for-agriculture training with CGIAR, the international farming-research partnership. The Nepal hub does crop monitoring, disaster damage assessment, and mapping for land rights.

The open part matters. The network says its hubs work with open-source tools, among them the open-source flight software common among hobbyist drone builders, and open mapmaking software that stitches a few hundred aerial photographs into a single field map. That is the network's claim, not an audit of every project, but the direction is real: capability that arrives as skills and readable code, not only as somebody else's black box.

The network's own decade of reporting counts thirty-eight labs and roughly seven hundred projects and trainings. The numbers are its own, self-reported. But what they count is not machines, and not acres: people who can fly, read the pictures, and keep the machines working in their own country. The sky over the world's farmland is filling with instruments. The interesting question was never whether the drones were coming. It is who flies them, and for whom.

The farm that owns no drone

Here is what most of the coverage misses. The farm that uses drones rarely owns one.

This should not surprise anyone who knows farm life. Farms have always rented big capability. The custom combining crew that sweeps through at harvest, the sprayer you hire rather than store: renting serious machinery by the acre is an old habit, because the economics of owning something you need twice a year have never made sense for most farms. The drone is simply the newest thing on that list, and the countries where drones are most normal have made the pattern official.

Start with China, because that is where the pattern is most developed, and where the evidence is clearest. Most agricultural drones there are not bought by farmers. They are bought by service companies and cooperatives, and the arrangements come in recognizable shapes. Manufacturers run their own service teams. An individual buys a drone and joins a local plant-protection team that handles training and dispatch and bids on government pest control contracts. Cooperatives buy their own drone and run it for their members: one machinery cooperative outside Nanjing fields a drone team that serves a hundred twenty farmers across some sixty-five hundred mu. And the newest shape is an app: dispatch by smartphone, the ride-hailing model, for a flying sprayer.

Canada's version looks different, but it rhymes. In Saskatoon, a company called Croptimistic has built a service around a camera that does not fly at all: it rides on the farm's sprayer during ordinary spray operations, photographing the crop every sixty feet. The company says it mapped more than a million acres of western Canadian farmland in 2025, its third season, and that its plant counts come within six percent of reality; both are the company's figures, unverified by anyone outside it, and the federal government was impressed enough to name it a national showcase example. What the farmer buys is not a camera, and not a drone: a map, a set of zones, a stream of advice, a way of seeing the field, delivered as a subscription.

There is a third Canadian shape, quieter than the other two, and for many farms it is the one that actually shows up: the adviser. The agronomist who adds a drone to the truck. The co-op agronomy desk that flies members' fields. The neighbour with a licence and a side business. The revolution, when it arrives at most farms, will not look like a robot in the barn. It will look like a familiar truck with a new case in the back.

That is the real shape of this revolution. The eye comes as a service.

Which means the question from the first chapter has followed us into the sky. When the eye is rented, part of the farm's sight lives in somebody else's building. Taranis runs its imagery through Google's cloud. Croptimistic's camera works only inside the company's own mapping service; you cannot take your fields to a rival and keep the history. Yaghi himself has a line for the moment the rented eye starts charging twice. On the Future Herd podcast, describing what equipment makers now bill for moving a farmer's own machinery records between machines, he said: "you're paying for the pizza, and then you're getting charged to share those slices to your friends. And that's not really building an ecosystem then." The pizza math applies to eyes too. None of this is sinister on a Tuesday morning. Services earn their keep, and these ones do. But it means the farm's ability to see itself has quietly become a dependency, and dependencies deserve to be noticed while they are still conveniences. This one comes back when we reach the fork.

The third way

There is another path, and it is my favourite story in this chapter.

Matt Reimer is a grain farmer in Manitoba. Some years back he wanted a grain cart that could drive itself alongside the combine, because the cart job eats a person's whole day at harvest. He could not buy such a thing, so he built one. He took open hardware and the open-source software that drone hobbyists use to fly their machines, taught himself programming through a free MIT course, and made his tractor autonomous. In one harvest, MIT's magazine reported, the driverless tractor came out to meet the combine more than five hundred times, and the arrangement saved him at least five thousand dollars that season.

The third way does not always look like one man and one machine. Donald Killorn is the executive director of the PEI Federation of Agriculture, the farmers' own membership organisation on Prince Edward Island, and his version of building is collective. On an island of small fields, no single farm is going to wire the province. So the federation goes up on the roof. His team was recently "on a roof in eastern PEI", he said, installing a wifi gateway, "basically creating the mesh network to bring the realtime data into our AI platform". Not a vendor's platform; the federation's. The field-sensor readings it gathers land on "a federation owned server", because, as he puts it, "farmers are very nervous about who's gonna own their data". He has told Future Herd listeners that he champions the farmers' organisation as "the trust layer in the data ecosystem": wherever there are farmers there are farm organisations, and they are the ones suited to bring the technology across to a thousand small fields.

The ask keeps growing. "My message has gone from farmers need to own the data, to farmers need to own the compute, to farmers need to own the electricity."

Same nervous system as the drone, different ownership. Rent the eye, buy the eye, or build the eye. A decade ago the third option barely existed. The tools Matt used were free to read, change, and share, maintained by a community of strangers who answer to nobody's shareholders. That combination comes back in the chapters about building your own tools. The sky is not only for rent.

What the view from above cannot see

Why does the view from above feel so much like truth?

There is a book I keep coming back to, Seeing Like a State, by the political scientist James C. Scott. It opens in a forest. In the eighteenth century, German scientific foresters looked at an old, tangled, mixed forest and found it unreadable: too many species, too many ages, impossible to count. So they remade it. One species, one age, straight rows, numbered plots. The forest became a table of timber, and you could manage it from a desk. The first generation of the simplified forest grew beautifully. Then the forest began to fail, because the table had left out almost everything that made a forest alive: the undergrowth, the fungi, the birds, the mess.

Scott's point is not that the view from above is useless. It is that the view from above is a simplification wearing the costume of completeness. Every map leaves things out, and the danger is never the map. The danger is forgetting that.

The drone's map is the farm made legible: rows, counts, stress scores, zones. It can see that a patch of field is struggling. It cannot see why. The why lives in the soil, in the history of that corner, in the thing the farmer's grandfather did with drainage sixty years ago, in knowledge that has never been photographed. The farmers who use the eye well all say some version of the same sentence: the drone tells you where to walk. Then the older eye takes over, the boot on the ground, the shovel in the soil, the hand on the leaf. The two eyes are not rivals. The skill of the programmable farm is the relay between them. Killorn states the rented eye's limit more precisely than I can: "we can outsource intelligence, but we can't outsource understanding."

One more thing about legibility, and then we can let the sky rest. A map made for you is a service. A map made of you is something else. The same flight that tells a farmer where to walk also tells the platform where the weak spots are, season after season, field after field. Scott aimed his lesson at states, but it travels well: whoever renders a world legible gets a say, at least a small one, in what that world is for. The antidote is what it has always been: the person on the ground, walking, keeping their own reading current.

Open questions

When the eye is rented, who learns? If the service company's agronomists see a thousand fields and the farmer sees one, where does the knowledge accumulate, and who owns it?

What changes in a farmer's own eye when the daily walk becomes a review of alerts? Is something quietly lost when the finding is done for you?

Who is allowed the overhead view of a farm? The farmer, the operator, the platform, the insurer, the neighbour, the state? Should a farmer be able to see everything the sky sees of them?

And if a kitchen in Toronto can tend a grove in Lebanon, what does local mean now? What does distance do to care, and what does it protect?

Agency moves

Look up. For one week, notice every aircraft that crosses your sky: the plane, the helicopter, the drone, the sprayer plane if you are lucky enough to live under one. Give each a guess about its errand. The sky over you has a work schedule you never get to read.

Find the free view. Open a free satellite map of a place you know well and study it for ten minutes. Then write down three things about the place that the view from above cannot tell you. Keep the list. It is a small working model of this whole chapter.

Ask one person who farms, or who advises farmers: did anything fly your fields this season, and what did it change? Farmers' market vendors count. The answer will usually surprise you, in one direction or the other.

Situation: The sky's to-do list

Two parts, one week, no equipment beyond a notebook and a phone.

Part one: the sky log. For seven days, every time you notice something overhead, write it down: the time, the place, the direction, and your guess at the errand. Passenger jet, news helicopter, delivery drone, crop sprayer, something you cannot identify. Do not research while you log; just notice. At the end of the week, read your log straight through. How much of the sky's work had you never noticed? Whose errands fill it? What would the log look like above a farm in August, or above your neighbourhood during a flood?

Part two: the two-eye exercise. Choose one place you know intimately: your street, your garden, a field you walk, a barn you have worked in. First, describe it from the ground, in five sentences, from memory and feet. Then find the same place on a free satellite map and describe it again, in five more sentences, from the eye in the sky. Now compare the two descriptions. What does the ground eye know that the sky eye never will? What does the sky eye see that the ground eye cannot?

That gap between your two descriptions is the argument of this book, felt in your own body. Every chapter from here on is about what a farm does inside that gap.

This one plays beautifully with a group. Same week, same sky, different logs; same place, different pairs of eyes. Compare the gaps, and notice that no two people describe the same place the same way from above, but everyone agrees on what only the ground can tell you.

The eye came down from the sky quickly. In the next chapter it goes somewhere smaller, warmer, and harder to look away from: the barn, where the cameras never sleep and the herd has started talking back.