iFixit vs the Factory: Who Really Owns a Repair Manual
iFixit has built 100,000 free repair manuals, and Valve's Steam Machine shows both the reach and the limit of that project.
Fifty-seven steps. That's the length of iFixit's guide to replacing the heatsink inside Valve's 2026 Steam Machine, and the internal power supply runs the same. Each step pairs a photograph with a sentence: this screw, this connector, this ribbon cable lifted straight up rather than sideways. Anyone who has opened a laptop on a kitchen table knows the shape of the problem these guides solve. A factory service manual assumes a bench, a torque driver and a part number already in stock. The person at the kitchen table has an hour, a credit card and no idea which of eleven identical black screws goes where.
- 1The library: iFixit hosts more than 100,000 free repair manuals, written by staff, students and volunteers, and edited by anyone who spots a mistake.
- 2The Valve test: iFixit published 18 repair guides for Valve's Steam Machine in September, three months after the console shipped, and the longest jobs run to 57 steps. It sells no parts for the machine.
- 3The standard: A public draft of a repairability standard developed with the National Sanitation Foundation took comments through October 9, an attempt to turn one company's scoring into something an industry can be held to.
IFixit has been answering that question since 2003 from a headquarters in San Luis Obispo, California, and it answers for free. The site hosts more than 100,000 repair manuals covering phones, laptops, cameras, cars and appliances, written by staff, students and volunteers, and editable by anyone who finds an error. That last detail is the entire design. A manual written by a manufacturer is a legal document with a liability clause attached. A manual written by a thousand people who have actually opened the device is a record of what breaks.
What a Wiki Does That a Factory Cannot
The mechanics of the model are worth spelling out because they explain why the library exists at all. iFixit's headcount sits around 268 people. Set that against 100,000 manuals and you get roughly 373 manuals per employee, a ratio no technical writing department could produce by hiring. The crowd writes the guides because the crowd owns the failures.
Failure patterns are local, physical and endlessly specific. A corrosion signature in a humid climate, a battery that swells after two years of overnight charging, a connector that cracks when a device lands on its corner. A manufacturer sees the aggregate of warranty returns as a line in a quarterly report. A technician in Nairobi or Manila or São Paulo sees the actual board, with the actual burn mark, and in much of the world the independent shop is the primary service channel because authorized parts networks are thin. A free manual carries different weight there than it does in a suburb twenty minutes from an Apple Store.
The wiki has real costs. Guide quality is uneven, coverage lags product launches, and a newcomer sometimes lands on a page for a 2014 model when they need the 2026 revision. iFixit's answer to that has been to build the missing scaffolding around the wiki: teardowns that show the inside of a device before anyone else has one, a repairability score that puts a number on the teardown, an app, and FixBot, a repair assistant that answers in plain language. The company's own framing is that it is building a free repair manual for every thing, which is a large claim dressed as a slogan.
Valve's Machine Makes the Gap Visible
The Steam Machine is an unusually clean test of what documentation alone can do. Valve's second attempt is a six inch cube running SteamOS on a semi-custom AMD Zen 4 six-core chip with RDNA-3 graphics, sold in 512GB and 2TB versions at $1,049 and $1,349. It shipped in June. iFixit put the full guide suite live on September 16, and unlike most of the library, these were written from Valve's own instructions rather than reverse engineered by a volunteer with a heat gun.
The list runs from an 11-step SSD swap to a 55-step motherboard job, with the heatsink and the internal power supply the longest at 57 steps each. That's a genuine map of a machine, published by the manufacturer through a third party, which would have been unthinkable from a console maker fifteen years ago.
Then the shape of the gap appears. iFixit doesn't sell replacement parts for the Steam Machine. Its store carries Steam Deck parts from the same partnership lineage, but for this machine the guides are live and the parts aren't listed. iFixit also hasn't assigned the Steam Machine a repairability score.
Read those three facts together and the architecture of the problem shows itself. Documentation is the cheapest thing a hardware company can hand over. A PDF costs nothing to publish, carries no warranty liability, and buys goodwill from exactly the customers most likely to complain. Parts logistics, serialized components and calibration software are where the money and the control actually sit. An eleven-step SSD swap is a repair the documentation enables on its own, because the buyer supplies the drive. A 57-step heatsink replacement without a purchasable heatsink is an education.
A Score That Wants to Become a Standard
IFixit has scored products for more than fifteen years, and the earliest version of its rubric makes the priorities explicit: design for repair carried 90 percent of the weight in v1.0. That's an honest description of what the score originally measured, which was how the object was built rather than what the company was willing to sell afterward.
The standard project changes the register. A score handed out by one organization is an opinion with a number attached, and manufacturers can ignore it or dispute it. A standard written through the National Sanitation Foundation's consensus process can be referenced in procurement contracts, warranty terms, insurer risk assessments and eventually regulation. The public draft went live this year and took comments through October 9, and the difference between those two artifacts is the difference between a review and a specification.
Crash testing is the useful comparison. Car buyers didn't adopt five-star ratings because the ratings were morally persuasive. They adopted them because fleet buyers, insurers and regulators started asking for the number, and the sheet metal changed to meet it. Repairability has no crash dummy, which is why the scoring question has stayed soft for so long.
IFixit has already shown how fragile its own numbers are. In 2023 the company retroactively dropped the iPhone's repairability score after parts pairing problems made the mechanical teardown look generous. The screws came out cleanly. The phone still refused to accept the result, and the score had to be revised downward after the fact.
The Parts Channel Sets the Ceiling
Right to Repair laws, whether the state statutes in the United States or the European Union's ecodesign requirements, are usually summarized as access to parts, tools and documentation. Two of those three are cheap. Every serious repair operation in the world can get a screwdriver.
The binding constraint is the parts channel plus the software that decides whether a fitted part counts as genuine. That's why the interesting line in the standard's own roadmap is that future criteria may take in parts availability and software updates, which is an acknowledgment that the current scoring captures the visible half of the machine.
When a guide and a parts listing disagree, check the device page's community answers before buying anything. The comment threads usually know which revision changed the connector, months before the guide is updated.
What a Photograph Cannot Prove
A photograph can verify a step. It cannot verify an authorization handshake. When a replacement battery triggers a warning or a display loses a feature after a swap, the repair succeeded mechanically and failed at the firmware layer, and there is no image of that failure to publish in a guide. This is the same verification problem that sits behind every AI-written answer a repair assistant produces: the text can be fluent and the device can still reject the work.
Parts pairing is the mechanism. A component that carries a serial number the mainboard checks against a server turns a $30 part into a $300 dealer visit, and none of that transaction shows up in a teardown. iFixit's scoring has tried to account for it since the iPhone revision, but the criteria keep chasing features that ship on a firmware schedule rather than a product cycle.
The Measurement Problem That Remains
Here is the difficulty nobody has solved. A repairability score is a snapshot of a machine on the day someone opened it, and a device's fixability can change eight months later with an update that re-pairs a component or a supply decision that stops stocking a board. No draft standard has proposed an interval for rescoring, and no manufacturer has volunteered to be measured twice.
The manuals will keep coming, because writing them is cheap and the community does the work. What the Steam Machine shows is that a free manual answers the question of how, and leaves untouched the question of whether anyone will sell you the thing the manual tells you to install. The score that eventually matters won't be the one that counts screws. It will be the one that counts what the company declines to put in a box.
Comments (0)
No comments yet. Be the first to share your thoughts.




