Two doors

The United States spent the year saying, out loud and for the record, that it would like to own Canada. Not a trade concession, not a border quarrel: the country. The farmers I know heard it the way you hear an engine pull into the yard at three in the morning and cannot tell yet whose it is. They did not file it under politics, because it never got withdrawn.

When I asked Tyler McCann of the Canadian Agri-Food Policy Institute about the year, he did not start with a tariff. He started with the ground: "those fundamentals that have been good for you for the last sixty years may not be there for you in the future." He called the risk generational, then put the fear in plain words: "what you have today is at risk." This chapter is where that warning stops being a forecast and lands on the table in front of a farmer.

On that table the whole reordering shows up as two pieces of paper describing the same week of March. The first is a renewal notice: software the farm has run for three seasons is billing again, the tier has moved, and the feature the farm actually uses now sits a level up, priced by people who have never stood in the field, due whether or not the rain arrives on schedule. The second is a booking sheet from a machinery cooperative of five neighbouring farms, names pencilled against a seeder and a sprayer, and a note about who holds the fuel card. One is a bill with an expiry date, on terms written somewhere else. The other is a schedule among neighbours, open to whoever shows up to argue about it. Both are about the same thing: who holds the machinery that grows the food. It is not a metaphor. It is decided on paper, at a table, every year, and for the first time in a long time the answer is not obvious.

The world the farm is plugged into

The choice is urgent now for a reason that has nothing to do with software.

For most of a century Canadian agriculture ran on one unstated assumption: that the country on the far side of the border would go on being a customer. Through 2025 the president of that country kept saying Canada was something else. He called it the fifty-first state, on the record and often, and described the border itself as an artificial line that would have to go. Canada's foreign minister answered in the plainest terms, that the aim was economic pressure until the country could be annexed, and the prime minister put it the way a farmer would: "There are some places that are never for sale." The tariffs that filled the news that year were the symptom. The cause was that the single largest buyer of what a Canadian farm grows had begun describing the farm itself as property.

The numbers were the visible half of it. When the tariffs landed in August 2026, Canada answered "dollar for dollar, rate for rate," and the counter-list reached into machinery parts, which meant the cost of fixing a machine became a trade question. China gave the other demonstration, colder and faster: a single anti-dumping ruling in August 2025 put a 75.8 per cent duty on canola seed and took the roof off the country's biggest export market for the crop, a February 2026 ruling drew it back partway, and a March suspension reopened some doors and left others shut. One crop went from a wall to a partial opening in seven months, and a farmer in Saskatchewan watched all of it from a kitchen with no vote in any of it.

The buyers were not the only thing that moved in 2026. The inputs a farm lives on are wired to the same unsettled world. Fertilizer follows whoever ships nitrogen and phosphate, and when Canada stripped Russia of its trade privileges in 2022 and China restricted its phosphate exports in early 2025, the cost of feeding a field went with them. Diesel rode a war in the Middle East that shut its shipping lanes, and by the 2026 harvest bulk farm diesel had climbed past two dollars a litre, roughly twice the price a year earlier, a change the Grain Growers of Canada worked out as about fifteen thousand dollars more to run one combine through one harvest. Then the weather put in its own entry, needing no buyer and no war: the 2026 harvest reached the prairies with the north too wet to finish on time and the south too dry, grain late where the water was, frost waiting behind it.

Underneath the buying and selling, the meaning of the word "tools" changed this year. The food and agriculture sector's own cyber threat-sharing body now counts state-sponsored hackers among the biggest dangers to it, second only to ransomware, and the actors it names are not trying to sell anything. Some are positioning inside the controllers that run grain handling and irrigation, and the doctrine behind it is not crime. In the military thinking of at least two major powers, disrupting another country's food system is a form of war, set down beside ports and pipelines. A farm's tools are now infrastructure in a conflict that has not declared itself, and the list of people who could turn off a Canadian farm got longer this year, with new names on it that send no renewal notice.

Tyler McCann put the practical version of what a farm can do with any of this: "We can handle tariffs as long as we have a tariff advantage." That is the reasoning of someone who has stopped expecting a stable rule and started thinking about position. His deeper worry was that Canada keeps exporting the raw and importing the finished: "we are losing that value added benefit and that value added advantage." He pointed out that thousands of Canadian food manufacturing jobs exist to serve the American market precisely because farmland cannot be moved south, and that those jobs, unlike the land, can move. John Barlow, the member of Parliament for Foothills, made the political version of it in the spring: "agriculture and food production as a whole needs to be elevated to a top priority in Ottawa."

The federal government has now said it itself. The National Food Security Strategy, launched in June 2026, states plainly that "our food chains rely too heavily on foreign suppliers." Its goals, in the strategy's own words, are "for more choice, more control, and more Canada," backed by more than $3 billion over ten years. Those words were written about processing capacity and grocery competition. They apply just as cleanly to software, and nobody in the document says so, which is the gap this chapter is about. A country that has decided to take back control of how it feeds itself, and still leaves the tools of that work on an annual licence from elsewhere, has made the largest choice on the list without noticing it.

What closed does best

Before the argument gets any further, the closed model deserves its strongest case, because it has one and the case is not stupid.

The first argument is access. In a Canadian farm paper, Mark Fournier, who teaches strategic planning at Olds College, asked the question "Is the future of farming billed monthly?" and answered part of it in the subscription's favour: subscriptions "can lower the cost of entry for farmers." Instead of buying a whole software package and then paying for upgrades that get delayed when money is tight, a farm can join a network and use the product while it pays for it. For a young operation with land payments and no spare capital, a monthly bill of a few hundred dollars to reach a technique that used to require a capital purchase is not a trap. It is the only door in.

The second argument is capability, and it is the strongest one. A subscription buys maintenance, security patches, model updates, and somebody to call. Open tools push that labour onto the user, and a farmer already has a job. When a piece of farm software is bad, the vendor is the only party with both the incentive and the ability to fix it, and a farm that stops paying stops receiving fixes. Anyone who has run an unmaintained open-source tool in a production setting knows the feeling of discovering that the person who understood it moved on three years ago.

The third argument is safety and compliance. Modern machinery carries emissions controls, safety interlocks, and autonomy features that have to be certified, and the manufacturer is the party answerable for them. A company that leases the guidance software and updates it every season is also a company that can be held responsible when an update goes wrong. This is the same argument the machinery industry made against easy repair, and it is not a lie there either. Part of what a closed system sells is an accountable party.

And the model keeps arriving in new places. Deere sells the intelligence rather than the iron now, and one American account of its strategy says so plainly: "The company sells subscriptions to the software that powers its automated sprayer, See & Spray." The pricing follows use rather than ownership, so the machine a farm buys comes with a capability it rents by the acre. That is a real product, doing real work in fields, sold to farmers who chose it.

What closed does when it closes

Here is the same model on a different day.

A farm management platform used across Europe, 365FarmNet, announced its own end: "On 30 November 2026, 365FarmNet will be discontinued." The notice is clear and even generous about the transition, telling users they can switch to the vendor's newer product, that "Your data can be migrated", that the basic licence will be free for a year, and that the migration runs through a dealer. Read that list again as a farmer and notice what has happened. A tool that held field records, machine data, and the shape of several seasons was switched off by a decision made in a boardroom, and the continuity of the farm's own records now depends on a migration process run by the company that is closing the door.

The migration is the good version, and it exists because a large owner found it worth doing. The pattern is less gentle where the owner is smaller. Food and agriculture technology has spent a decade full of platforms that raised money, signed up users, and then stopped existing, and the collapse is worse for the user than the shutdown, because a shutdown comes with a notice and a bankruptcy does not. Growers attached to a container-farming company that filed for bankruptcy discovered that the machines they owned needed a company that no longer answered. A platform with a million registered farmers can end for reasons that have nothing to do with those farmers.

There is also a quieter version that never makes news, and it is the one that should worry a working farm more than any shutdown. Telematics hardware depends on the cellular network it was built for. When the carriers retired the older generation of that network, machines on the farm stopped reporting, no rule changed, no company failed, and the fix was to buy newer modems. The data did not stop because anyone decided to take it. It stopped because the infrastructure underneath it was retired on a schedule the farm did not set and could not see.

Now put all of that beside the renewal notice on the kitchen table, and the clause that matters is not the price. It is the calendar. A subscription does not only meter a tool; it holds a season. The farmer does not choose when to renegotiate, because the renewal date sits in whatever month the vendor picked, and a decision to switch tools has to fit inside a window that planting and harvest have already booked. That is what it means to say the subscription owns the calendar: not that it sends a bill in a bad month, but that the farm's ability to change course is fenced into the vendor's paperwork.

What open does best

The other door is older than software, and in Canada it has a name and a history.

Machinery cooperatives are common in Quebec and Ontario, and their value sits in mechanics so plain they are easy to dismiss. A CUMA, the French acronym that stuck, owns equipment on behalf of its members, each of whom has one vote in the operation of the co-op regardless of how much capital they have invested. That rule is a constitutional fact, not a nicety: the farm that put in $400,000 of the combine does not get more say than the farm that put in $40,000. And the economics are set the same way: the co-op operates at cost, returning any surplus after expenses to the members in proportion to their use. No margin is being collected by anyone, because the entity collecting it is the members.

This is not a software-era invention, and the Prairies have their own version. In the late 1930s and 1940s, farmers in Western Canada built Canadian Co-operative Implements Limited, a company whose entire promise sits in the title of the history written about it decades later: "Better Tractors for Less Money". The machinery dealers of the day set prices a farm could not argue with, and the only move left was to become the dealer. When the terms of a tool are set elsewhere and cannot be negotiated, farmers have historically done the one thing that works: acquired the tool collectively.

Software has a version of that too, and it is running. Open Food Network Canada is marketplace software that "removes the barrier to entry in the marketplace found in commercial software." It is operated by a non-profit rather than by a company with a growth target. On the machine side, the guidance software that a farmer in central Alberta wrote for his own operation, and then released to whoever wanted it, is still out there being installed on equipment that the factory option would have priced at a different order of magnitude.

None of that is new, but one thing about it changed, and recently. For most of the software era, the freedom an open licence grants, the right to read the code and change it, was a freedom most farmers could not use, because changing code meant hiring a programmer. That gate is open now. A farmer can describe the tool he needs in plain language, an AI writes it, and when something is wrong he says so and it changes; the awkward name for this is vibe coding, and this book gives it a chapter of its own. When the ability to modify a tool stopped being rare and expensive, the value of owning the source grew by the size of the gap between a tool you can rewrite and a tool you can only rent. Software that acts rather than only answering, the agentic kind, widens that same gap, because inside a system you are allowed to read there is no ceiling, and inside a closed one it stops at the button the vendor built.

The design point is not that these tools are free. It is that their rules are readable. A cooperative's members can see the booking sheet, argue about it at a meeting, and change it by vote. An open-source project's users can read the code, pay someone else to improve it, or leave with a copy. Both arrangements make the terms visible, and visible terms can be argued with. That is the whole of the advantage, and it is bigger than it sounds, because a term you cannot see is a term you cannot contest.

What open does when it fails

The other door has its own failures, and this chapter owes them the same treatment it gave the first one.

The first is maintenance. Open source is a licence, not a workforce, and a project without a maintainer is a liability wearing a permission slip. The research behind this book found a well-documented open data standard for farm machinery with a working reference server and zero third-party commercial adoption, which is a polite way of saying that everybody agreed it was a good idea and nobody shipped it. Most community software has no support line, no warranty, and no obligation to anyone. When it breaks in the middle of a season, the person who fixes it is the farmer, at night, with a forum open in another window.

The second failure is concentration, and Canada has an unusually clear case of it. Co-op Atlantic was a federation of co-operatives across the Atlantic provinces, and in 2015 its members voted in favour of selling the food and gas wholesale and retail business to Sobeys. The decision was made properly, by the people who owned it, ratified by representatives of around sixty co-ops. Sobeys then rebranded the stores. Nobody stole anything. A co-operative is a business, businesses carry debt, and members facing a hard balance sheet can decide that the strongest version of their future runs under somebody else's banner. That is the risk in the open side that no licence prevents: collectively-owned institutions are still institutions, and they can be sold, wound down, or absorbed, with every vote correctly counted.

The third failure is a boundary on what gets built openly at all. The research found that where liability and capital concentrate, open hardware tends to be missing, and the terrain is occupied by subscription products instead, including virtual fencing, where a handful of firms hold the ground and the collar only works while the bill is paid. Nobody is going to build an open, certified, liability-bearing autonomous machine in a garage, and pretending otherwise would be a fairy tale in a book that has spent this long insisting on evidence. Openness has conditions of success, and in some parts of farming the conditions are absent.

Who could turn off your tools

The Situation at the end of this chapter asks the reader to write a list of everyone who could turn off their tools. It is worth doing here first, in plain form, because the list is longer than most people expect and its length is the actual argument.

A platform vendor can discontinue a product, as 365FarmNet is doing. A network operator can retire the generation of hardware your equipment was built for. A dealer can stop stocking a part or stop supporting a machine older than its service window. A software company can change a licence tier and move the feature you use one level up. A cloud provider can change its terms, raise its price, or lose interest in the market. A standards body can revise a specification so that a working tool no longer talks to anything. A lender can take the equipment a farm put up as security. A company can go bankrupt, as farm technology companies do at a rate that would be alarming elsewhere. A government can change a tariff or a rule and alter what the whole operation is worth, which is the lesson the canola growers got in a single August. And now a state can reach the machines themselves, not to send a bill but to stop food moving, which is what the threat reports mean when they say the sector is being positioned for conflict.

Some of these are unavoidable, and a farm that tries to own everything it depends on will run out of money before it runs out of list. That is the frame. The point is not to eliminate dependence. It is to know the shape of the dependence, because the difference between a farm that can adapt and one that cannot is usually the difference between a farmer who knows exactly which door can be closed on them and one who learns it at the worst possible time.

The choice is a portfolio, not an identity

Which brings the chapter back to the kitchen table, and away from the version of this argument that turns into a personality test.

Douglas Rushkoff's formulation is the useful one here: program or be programmed. It reads as a challenge, but in practice it is a description of what is already happening at every level of the system. Someone decides how the guidance algorithm weights a wet spot in a field. Someone decides which detections a sprayer will act on. Someone decides whether the app tells the farmer the machine needs one pass or three. A farm that has no idea who those someones are, or what they optimised for, is programmed. A farm that knows, and can walk to the next door when the answer is bad, is not.

The same history runs underneath this, which is where Silvia Federici's account of enclosure does its work in the book. The commons were fenced once in land, by processes most of us know as the early modern history of agriculture, and the operation repeated whenever a shared resource became valuable enough to privatise. Software is the newest field under that plough, and the fencing does not look like a fence. It looks like a licence tier, a proprietary port, a service that ends, and a clause about who may read your data. Farmers who have watched a shared resource become a subscription have already lived through this once, and they know how the story ends if nobody argues.

So the fork is not a moral identity, and framing it as one is how it gets lost. It is a portfolio decision, made tool by tool and season by season, and the questions are mundane. Does this thing do enough to be worth renting. Can I read the terms. What happens if it stops. Is there a second door. A farm that can only answer those questions after the fact is running an experiment with its own season as the subject.

The open craft and the closed industry will both exist, and most farms will use both. The one whose terms the farmer can see, argue with, leave, and rebuild is the one that survives a bad decade, and a bad decade is what the year just brought to the door, by tariff and by a neighbour who keeps saying he wants to own the country.

Open questions

Where does the fork actually land? The intro frames the choice and this chapter stages it, which was the shape settled before drafting. What this chapter adds is the argument that the events of 2025 and 2026 changed the stakes: a farm whose prices are set by trade decisions and whose tools are held on licence is exposed in two directions at once, and the two exposures compound.

Can the closed model be described fairly enough to be worth choosing? It can be described better than it usually is, which is what this chapter tried to do, and the interesting question is whether a farm can take the subscription's advantage in capability while keeping an exit. Vendors do not generally sell the exit.

What does an open tool owe its users? Open source promises the freedom to fork, which is a promise about permission rather than about maintenance. Cooperative ownership promises governance, which is a promise about who decides. Neither promise answers the question of who shows up when the machine breaks on a Saturday, and most of the disappointment in this space lives in that gap.

Does Canadian policy treat farm software as infrastructure? The National Food Security Strategy wants more control over the food system and treats processing capacity and grocery competition as the levers. Whether the tooling layer gets the same treatment, in procurement, in data policy, in what the federal government chooses to run on, is open, and the answer will be decided by decisions that are not currently described as agricultural policy.

Moves you can make

Write the list. Name every entity that could turn off a tool you depend on: the software vendor, the dealer, the network operator, the cloud service, the lender, the platform that might get bought, the government whose rules decide what your harvest is worth. Then mark the ones you could replace inside a season, and the ones you could not. The second list is the one to keep somewhere you will see it.

Pick one item on that list this month and find its open alternative. Not to switch, to know the exit exists: a machinery cooperative or a tool-sharing arrangement in your area, a farm management project maintained by a community or a university, a marketplace run by a non-profit, a farm equipment community that publishes its own designs and answers questions, a local shop that does the unofficial version of a job. Twenty minutes of looking turns an abstract dependency into a named option, and a named option is worth a surprising amount the day the first one closes.

Ask three questions before the next renewal. What happens to my data if I stop paying. What happens to this product if you stop making it. What can I export and use somewhere else. Ask in writing and keep the answer, because that answer is the actual contract.

The situation: the renewal notice

Take this season's renewal notice for whatever software you already depend on, and before you pay it, fill in four lines. What this tool actually does for me, in plain words. What would happen if it stopped tomorrow. How long it would take me to replace it. What the second door is, by name.

Then decide with the paper in front of you. Pay it, if the tool has earned it and the exit is real, because a tool that has earned its price is a good outcome. Or spend one evening finding the alternative before the renewal date arrives, because the whole difference between the two doors is that one of them you can walk through on your own schedule.

What the farm does with the machine is one question. Who holds the record of what the machine did is another, and it is the one that decides how much the first question is worth. The data leaves the field, and what happens to it next is where this book goes from here.