🔧 Nobody Graded the Homework September 7, 2026 · https://tavi-blog.github.io/nobody-graded-the-homework/

I had a spec sheet and a repairability score open in two tabs last week, doing the same comparison I've done for every phone I've seriously considered for years, and the field where the actual repair manual was supposed to live just wasn't there. Not hidden, not vague. Empty. I assumed it was a bad link on one retailer's page until I went looking and found out this is close to the norm right now, not the exception.

The EU has required a self-reported repairability score on smartphones and tablets sold there for about a year, covering things like how easy the device is to open, whether parts are available, whether fasteners require special tools. Alongside the score, manufacturers are supposed to publish where a buyer or an independent shop can actually go to get the repair instructions and spare parts pricing that the score is claiming to summarize. A report out this week checked several thousand of those listings against that second requirement and found that only about one in five of them actually pointed to a real page with real repair information behind it. The rest is a number with nothing under it, a grade with no attached homework to check it against.

What actually bothers me here isn't the compliance rate, bad as it is. The score is self-reported to begin with, and nobody with regulatory authority has apparently gone back to check whether the reported numbers match reality. A manufacturer fills in its own repairability rating the same way it fills in a nutrition label, except there's no equivalent of a food inspector who occasionally walks in and weighs the actual product against the box. The system assumes good faith from exactly the parties who have every incentive to round a fastener rating up or leave an inconvenient field blank rather than admit a phone is glued shut.

I want to be fair to why it's built this way, because the alternative isn't free. A verification regime, third-party labs pulling phones apart to confirm every claimed score, would have taken years longer to stand up and would have applied to a much smaller slice of the market on day one. Regulators had a real choice between a self-reported system that covers every phone sold in the bloc immediately and an audited one that covers a fraction of them a few years from now. I'd probably have made the same trade. A repairability score that exists on paper, even an unverified one, at least creates the expectation that repairability is a thing worth disclosing, and that expectation didn't exist before. The problem isn't that they chose speed over rigor. It's that a year in, nobody has come back to add the rigor, and a rule that only works on trust needs someone checking whether the trust was earned.

This lands somewhere specific for me because the actual mechanics of my job involve exactly this problem, just in a different industry: numbers that get entered by whoever benefits from them looking good, sitting in a system nobody audits until something breaks downstream. A field that's supposed to hold a real answer sits blank often enough that you stop being surprised by it, and you learn fast that a dashboard full of green checkmarks isn't the same claim as a dashboard someone actually verified. That instinct is exactly what I brought to two browser tabs and a blank repair-manual field. I wasn't shocked. I recognized the shape of the failure before I understood the details of the policy.

So the two-part test I've run on every phone purchase for years, weighing labor and privacy on one side against whether the thing actually works well enough to live with on the other, just picked up a complication I didn't ask for. I used to treat a published repairability score as a fact I could weigh against the rest. Now it's a claim, made by the one party with a reason to shade it, that I have to treat the way I'd treat any other unaudited number: useful as a rough signal, useless as a promise. I still don't know if the phone I'm looking at actually opens the way its score says it does. I just know a lot more about why the label exists and a lot less about why I should trust it.