The repair shop

Every farm has a building where the machines get fixed, and it is usually the least photogenic building on the property. A bench with a vise bolted to it. A shelf of filters and belts. A coffee can of bolts kept for thirty years and needed twice. When something comes in wrong, the work starts the way it always has: listen, look, take the cover off, find what failed, put it right. There is a newer tool in that room now, in the chest pocket of whoever is standing over the bench: a phone that will fetch a manual for nearly anything and answer a question in plain sentences at two in the morning.

This is the skill a farm actually runs on. Seeds, feed, fuel, and labour are all purchases. Repair is the one that compounds, because every part a farmer can diagnose is a part they do not have to buy again, and every hour of repair is an hour the machine spends working instead of waiting.

Then the day arrives when a machine stops mid-season, the screen lights up, and the message is not a symptom. It is a number. The farmer has rebuilt engines. He cannot read this. The machine knows what is wrong, stores the answer in a controller, and will only share that answer with a diagnostic program he is not allowed to have. So he phones the dealer, the service truck comes when it comes, the technician plugs in, reads the code, clears it, and bills for the visit and the travel. In a good week that is an inconvenience. In the first week of harvest it is a field standing ripe in front of rain.

Nothing about that sequence is natural. The machine can tell him exactly what is wrong. The decision not to let it is an arrangement, made by people and written into software. A machine you cannot fix is a machine you rent, whether or not the paperwork uses that word.

What the lock is made of

The National Farmers Union puts the point in a single sentence on its right to repair page: "Fix your own machine and your warranty is void." The union defines repair plainly: "The right to repair means goods, owned or leased, are repairable at a reasonable price, within a reasonable time-frame by the owner or a repair service of the owner's choice."

What is locked is more specific than the word suggests, and the clearest public inventory of it was written by regulators. In the settlement the Federal Trade Commission and five states reached with Deere in July 2026, the company agreed to make repair resources available to farmers and independent repair shops on the same terms as its dealers, and the order lists them one by one: "Reading, clearing and resetting electronic fault codes"; "Reprogramming of electronic components"; restarting a machine after an emissions-related shutdown, which the order says is "commonly referred to as" limp mode; and "viewing and searching technical manuals" along with troubleshooting solutions.

Read that list as a farmer rather than as a lawyer and the shape of the problem appears. The machine's own knowledge is the first lock, since fault codes sit behind software only dealers have. Pairing is a second and stranger one: a replacement part is not finished when you bolt it on, because the machine has to be told the part is legitimate and only authorised software can tell it. Limp mode is a third, where the machine protects its emissions system by refusing to work properly until somebody with the right tool resets it. Manuals are a fourth, since the document that would let an independent mechanic do the job correctly sits behind a licence. None of these are physics. Each is a decision that can be made differently, and in some companies is.

The makers have a case, and it deserves their own words rather than a caricature. Interviewed about a Canadian right-to-repair bill, the president of the North American Equipment Dealers Association argued that "Customers are able to access all the tools they need to repair their own equipment," and warned that "Amending the copyright act does bring in a whole bunch of unintended consequences." The interoperability fight draws the same defence: digital locks, in the trade framing, protect copyrighted technology and prevent hacking. The deeper argument is about liability. A modern tractor is a tonne-class machine moving at speed, its engine has to meet emissions rules the manufacturer certifies, and its failure modes include things that kill people. Nobody who builds those machines wants firmware they cannot vouch for loaded onto a combine with a person in the seat. In Europe the equipment-makers' federation warns about laws "designed without understanding rural realities."

Both things are true at once. The safety argument is real, and the same infrastructure that protects a driver's life also protects a dealer's revenue, because the tool that prevents an unsafe modification is the tool that prevents a competitor from quoting the repair. That is the ordinary shape of a lock: a legitimate purpose with a commercial consequence bolted to it, and a company that gets to decide which one it is talking about on any given Tuesday.

What Canada changed

For most of the last decade the legal advice to a Canadian farmer was that opening a digital lock, even to fix your own machine, could put you on the wrong side of the Copyright Act. That changed in the fall of 2024, when Parliament passed two amendments making it legal to circumvent what the law calls a technological protection measure in order to diagnose, maintain, or repair a product. A separate amendment to the Competition Act takes aim at manufacturers who refuse to supply the parts, tools, or software a repair needs.

What the amendments did not do matters more than what they did. A right to open a lock is not a right to receive a fault code, a service manual, or a diagnostic program. Canada made it legal to open the lock. It did not make anyone hand over the key.

The provinces have been the more concrete level, and each has taken a different corner of the problem. Rob Henderson, a Prince Edward Island MLA, put the stakes in the least abstract way available: a broken washing machine delays your laundry, while broken farm machinery "can affect your livelihood in a major way." Quebec has a planned obsolescence law, Ontario a right-to-repair bill that names agricultural equipment outright, and Saskatchewan has been licensing machinery dealers since before anyone put software in a tractor. The makers meet each of these the way they met the federal bills, arguing that the rules were written for electronics and appliances and that farming should be carved out.

What the lock costs

Downtime is the price that never appears on a repair invoice. A machine that stops in the seeding window or the harvest window loses the best days of the year, and those days cannot be bought back. In Prince Edward Island the federation representing farmers made the environmental version of the argument, and its executive director, Donald Killorn, put the real hierarchy of green choices in a sentence: "The things with the heavy metals and valuable minerals in them, being able to extend the life of those will be key if we want to become a more sustainable society." A machine that can be repaired stays in service, and a machine that cannot is scrapped early and replaced, with the energy and mining of the replacement embedded in the decision. Buying a new tractor is the least sustainable way to solve a fault code.

The industry economics point the other way. Farm implement manufacturing in Canada is worth roughly $2.4 billion in exports and $8 billion in annual revenue, according to the Agricultural Manufacturers of Canada, and the implement makers say incompatibility is spreading. Interviewed by CBC, the general manager of Honey Bee Manufacturing, a Saskatchewan company that makes combine headers and employs about two hundred people, said that incompatibility is "becoming more and more prevalent every day and every year". A company that builds a header has to prove it will work on a machine somebody else built, and every proprietary port makes that harder.

On the farm the arithmetic is more brutal. A repair the farmer cannot do is a repair that waits for a schedule, and scheduling has nothing to do with weather. The alternative is to own two of everything, the insurance policy that turns a working farm into a capital-intensive one and decides who stays in business. It also resets what ownership means, because an asset you cannot service on your own terms is closer to a subscription with a title deed. That is the sentence the National Farmers Union has been pushing at for years, along with its sharper version of the politics, that this is an issue which "pits farmers against multinational machinery companies that use standardization, compliance, and control over farm activities to make more money."

The assistant in your pocket

Here is where the last two years change the texture of this problem, though not in the direction the marketing suggests.

Start with what a farmer already does. In February 2026, Deere's own farmer publication told the story of a Kansas farmer who needed to work out whether he could run two trucks at harvest instead of three, and who instead of building a spreadsheet put the question to ChatGPT. "ChatGPT," he said, "probably took a minute to solve." The same article lists what these assistants are used for around him: diagnosing repairs, running calculations, working through problems that used to belong to a specialist. The tool is general, it costs almost nothing, and it answers in plain language.

Then there is the version built by people who fix things for a living. iFixit, which has spent two decades assembling the largest public repair library in the world, released an assistant called FixBot that works from its own documentation instead of the open internet: "a calm, knowledgeable expert that helps you figure out what's actually wrong and find the right documentation." It draws on more than a hundred and twenty-five thousand repair guides plus a very large question-and-answer archive, and its stated advantage over the general tools is that it is "much better than other systems at not making stuff up." That is the property that matters when a wrong guess means a burned hydraulic pump or a hand in the wrong place. It works beautifully on a laptop with a swollen battery. It cannot work on a tractor whose fault history is behind a dealer's licence, because there is nothing for it to read.

That boundary is the argument. An assistant of this kind can only reason about what its user can see, describe, or find, so its value is set by the openness of the machine it is helping with. Where the code is visible, a farmer with a phone has a second opinion available at two in the morning, and the second opinion is often good. Where the code is hidden, the same phone is useless, and the assistant is working on the other side of the wall.

Which is why the vendors are building their own. At the 2026 Farm Progress Show, Deere revealed an assistant it calls JD, described in trade coverage as helping farmers find information and analyse data through the systems they already use. Its service product, Operations Center PRO Service, now walks a farmer through repairs step by step, with an AI feature that searches equipment manuals. Consider the combination: an assistant bolted to the platform that holds the machine's data, sold by the company that controls the fault codes. On the dealer's side of the same fence, service businesses pitch software agents that answer parts calls around the clock, triage fault codes, and dispatch mobile service trucks during harvest. Each is a real convenience, and each deepens the arrangement this chapter describes, because the assistant that knows the machine best belongs to the company that decides who else may know it.

None of this makes the farmer's assistant pointless. It makes its value a function of access. The tool in your pocket is not a crowbar. It is a colleague, and colleagues are only useful in a room they are allowed into.

The people who fix it anyway

Farmers have never been passive about this, and the most interesting response is not a lawsuit. In central Alberta, a grain farmer named Brian Tischler ran into a problem with field mapping and section control on his own equipment, and rather than buy his way out of it he wrote the software. AgOpenGPS started as one farmer's fix, went out through agricultural forums, and was released on GitHub, where it grew into a public project with hardware designs, firmware, and a community that a third-party account puts at more than 2,500 members. What it delivers is guidance, autosteering, and mapping a farmer can install on the machine already in the yard, at a fraction of what the factory option costs. It is not a company. It is a group of people who decided the answer was to own that layer themselves.

Around that project sits an informal economy of repair the official systems were not built to notice: the independent mechanic with a laptop and a willingness to read forums, the used-parts network, the neighbour with the same machine who has already been through this, the farm-supply counter that knows which aftermarket part fits. The habit is old. Farmers have built their own tooling and workarounds for as long as they have farmed, and the software lock is the first obstacle in a long time that cannot be solved with a welder.

The evidence that is not there

The practice is everywhere and the record is missing. Farmers describe using an AI assistant in the shop. Trade publications describe it, vendors market it, and it is happening in kitchens and Quonsets across the country, where someone holds a phone in one hand and a flashlight in the other and asks what a noise means. What does not exist is a record: no dataset of assisted farm repairs, no measured success rate, no published account of where the advice stopped being useful. The stories are real and the documentation is absent, and the absence is not the farmers' fault. Nobody is paying for that research, and the companies holding the fault codes have no reason to want it run.

The way that gap gets closed is boring and available to anyone. Write it down. What the machine said. What you could see without the tool. What you asked, and what came back. What you tried, and what worked. A page like that is a small inconvenience in a bad week, and across a season it becomes a record of a practice the industry currently describes only in marketing. Enough of those pages become evidence, and evidence is what the policy debates in this chapter have been missing.

Open questions

Does legal permission matter without access? Canada now permits a farmer to circumvent a lock to repair a machine, and a right that requires hardware, software, or a password nobody will give you is a right that exists mainly on paper. The next round of this argument is about obligations rather than permissions: whether anyone must be required to hand over the tools, the parts, and the documentation that make the permission usable.

What would verification look like? A repair log kept by a farmer is a start, but a claim about a method needs more than one farm's anecdote. Independent shops, dealer technicians, and researchers could measure the same thing with one shared protocol: what was diagnosed, with what information, at what cost, and how often the first answer was right. Nobody has built it yet.

Who owns the fault history once an assistant reads it? The moment a farmer describes a machine's symptoms to a cloud service, that description leaves the farm, and the same is true of a photograph of a serial plate or a scan of an error screen. The manufacturers already hold the machine's telemetry. The interesting question is what a farmer's own query history becomes, and to whom it is worth something.

Does the right to repair extend to what is inside the machine? A part that must be paired is a part you cannot buy from anyone else, which makes the pairing requirement a competition question wearing a technical costume. The most useful regulatory question about repair in the coming years may turn out to be a question about parts.

Can a farmer take the productivity of the new machines without the dependency? The machines are genuinely better: autosteer, section control, and precise application save fuel and inputs, and the case for buying them is real. Whether that case can be separated from the subscription, the dealer's monopoly on diagnostics, and the telemetry flowing back to the maker is the question the next farm generation will spend real money answering.

Moves you can make

Name three things you own but cannot fully open. Not things that are broken, things that are closable: the phone with a battery you cannot replace, the appliance whose error code leads to a service line, the machine whose diagnostics need a laptop you do not have, the account whose data you can download but not take anywhere. Write the three down, and for each one write the name of the thing standing between you and the inside.

Find out who is working on the third one. Every locked object has a community around it somewhere: a forum, a repair guide, a GitHub project, a local shop that has been doing the unofficial version of the job for years, a maker space, a right-to-repair group. Spend twenty minutes finding one and reading what they know. This is how the skill in the shop got transferred for a hundred years, and it still works.

Start the log. The next time something breaks, write down four things before calling anyone: what the machine said, what you can see or test yourself, who else could look at it, and what it cost you in hours. Then, if an assistant helped you think it through, write down where the advice stopped being useful. Keep it for a season. That page is worth more than this chapter.

The situation: find the wall

Pick one thing you own that you could not open or fix if it broke tomorrow. It does not have to be a machine: a phone with a battery you cannot reach, a laptop that runs hot, a bicycle hub, a lid that will not come off, a car, the baler in the barn. Then spend a single afternoon finding the wall.

Start with the phone. Ask it what is wrong and how to open it, and write down the exact sentence where the help stops, the one where it says it cannot see what you are looking at, or that you need a tool, a code, or a login. Then put the phone down and see what you can work out with your own eyes and hands before you have spent a dollar. Then find the person or the place that actually could fix it or open it, the one the phone or a forum pointed you toward, and write down the one thing they have that you do not.

The afternoon produces a page, and the page is not a repair. It is a wall, described: the thing, the sentence where the help stopped, and the single missing piece on the other side of it. What the page usually shows is that your ownership stopped earlier than you thought, at a screw that is not there or a code that belongs to somebody else. The next time something breaks, or refuses to open, you will already know the shape of the room you are not allowed into.

The alternative to a farm that can repair its own machines is a farm that waits for permission, and the two of them are not the same business. Which is the subject of the next chapter: two doors, and the question of who holds the keys to each.