Well then! Almost a full year since I last published anything on this little nook of the internet. Last I had written about becoming a physics teacher, and that's the best excuse for what's been keeping me away from updating my personal site for the longest time. Teaching has been fantastic, and while I won't get into the details of that, I'll just say that it's easily been the most rewarding work that I've ever done. I would say that it's taken me a good 1.5 years in finding my routines and establishing my educational style, which is more about classroom management than it is about rigor, per se. But I digress.
What follows is less from my seat as a teacher than as a software engineer who happens to teach. I want to talk about what the computer is actually doing in the American public classroom — what it was supposed to be doing, what it's doing instead, and what it's quietly taking with it on the way out.
Claim The modern public education system relies heavily on technology to outwardly appear to have modernized and meaningfully improved the quality of education that public servants of the state and community, our teachers, are able to provide. Every high school student of the district I am employed in is given a Google Chromebook (this device in and of itself is a problem that I will return to), that they are expected to keep with them for the four-year secondary (high school) education cycle. As educators, we are strongly encouraged to deliver pedagogy via the vehicle of technology, so much so that "use of technology" is one of many metrics for evaluating educators during mandatory observations. "Technology use" here appears to be rather loosely defined (or interpreted by admin) to include: showing slides on a ViewBoard, using a ViewBoard as a digital whiteboard, distributing educational content to their Chromebook, and requiring that students use the EdTech applications that a given district has paid out large contracts for.
This has been the trend in US public education since at least 2015, when Chromebooks first crossed 49% of K-12 device shipments in the United States and "technologically literate" became an undercurrent of what a modern education was supposed to provide. The framing was, and is, sincerely held. The procurement was not based on it. The decisionmakers in district offices were optimizing for cost, IT manageability, and the politically appealing image of a laptop in every backpack — not for whatever long-term cognitive effects a decade inside a browser-based operating system might have on a fourteen-year-old.
We collectively decided, somewhere along the way, that everyone needs to be good at using computers. That was a reasonable inference from forty years of personal-computer productivity gains in the white-collar workforce, and it would have been a fine policy if the computers we put in classrooms were the ones the productive part of the workforce actually uses. They aren't. There are roughly two timelines this could have gone down, and we ended up in the worse one. Call ours the lower-standard timeline: an ad-and-cloud company won the K-12 contract by giving away the productivity suite for free, shipping a stripped-down browser OS on commodity hardware, and rebranding the resulting kiosks as "computers." US K-12 standardized on Chromebooks within a few years; by 2016, 58% of school-purchased mobile devices in US K-12 were Chromebooks, up from 50% the previous year.
It isn't a conspiracy; it's an incentive alignment that nobody on the public-sector side had the technical literacy to push back on. Google's business is routing user attention through cloud services to support ad revenue and SaaS contracts. A web-only operating system shipped on cheap commodity hardware solves that for them at industrial scale: the device requires a connection, the connection requires an account, the account lives inside the productivity suite. The district gets a $250 laptop and a free email/docs/classroom stack, with a managed-device console that lets a single sysadmin wipe and reissue a hundred machines in an afternoon. In exchange, an entire generation of students learns what the word "computer" means inside Google's walled garden. That isn't malice. It's what happens when one side is running a business and the other side is filling out a purchase order.
In this lower-standard timeline, and in my observations, computer literacy has actually decreased. By using these "computers" that are dependent on the cloud, in tandem with over-dependence (closer to addiction, according to evidence) on smartphones, we have abstracted away many of the paradigms that created proper computer literacy. By creating data abstractions that let users throw things into a digital storage "cloud" (Google Drive for Chromebooks, iCloud for iPhones), we've produced a cohort that no longer reliably understands basic file-hierarchy organization. This is by now well-documented by educators: an astrophysics professor at Cal Poly Humboldt, Catherine Garland, kept asking students where on the computer they'd saved a file — desktop? shared drive? — and getting blank stares, because the question didn't parse for them. Many of them did not know where their files lived, and more importantly, did not understand the question. This particular point could be seen in two ways. The first viewpoint would be that file based heirarchy is not the optimal organizational structure for humans, and that we have become over-familiar to the UNIX "everything" as a file philosophy. In other words, we became too computer like, and perhaps being computer-like is not of value (Turing and modern LLM architecture would disagree). I suppose you could say from this view there is more opportunity in formalizing how we organize information as humans, which is an interesting philosophy for another time. The second, the view cloud providers prefer, is that users shouldn't have to think about organizing their data — that built-in search is good enough, and the cognitive labor of filing is wasted effort. This pushes users toward a flat structure: one big bucket, maybe a handful of broad folders by topic, and a search bar to find anything specific. No nesting, no hierarchy, no decisions about what belongs with what.
The catch is that the act of organizing is the thing that trains you to organize. Deciding whether a document belongs in "physics" or "physics/labs" or "physics/labs/spring-2024" — or noticing that the document doesn't fit any of those and forces you to invent a new category — is a small but real cognitive exercise. You're building a mental map of how your information relates to itself. Skip the practice and you lose the skill. Worse, a flat structure imposes a ceiling on how complex your organization is allowed to get: there is nowhere for nuance to go. So users converge toward fewer, broader, vaguer buckets over time. Their organizational vocabulary shrinks to fit the tool.
This is the same dynamic, in miniature, as the contemporary worry about offloading thinking to AI. The muscle that doesn't get used atrophies. By the time a student reaches a college lab notebook or a professional codebase, they've spent a decade in an environment that actively discouraged them from building a hierarchy of anything. And the smartphone — where the data attached to any given app is almost entirely abstracted away from the user — makes Google Drive look hierarchical by comparison.
So what does the higher-standard timeline look like? Less radical than you'd think. The computers we put in classrooms are simply the ones the professional world uses — a laptop running a real operating system, with a filesystem you can see and a terminal you can open if you want to. Open Finder on a Mac or Explorer on Windows, switch to list view, and you can see exactly what's happening: parent folders contain children, children belong to parents, all the way down. That isn't advanced. It's how every computer works under the hood. The only difference is that you haven't been kept away from it.
Everyone in this timeline comes up with their own organization, the same way they organize their own bedroom or their own closet. Organization is the process of deciding where things go; putting things back where they belong according to a system someone has already set up is a different activity, and is called cleaning.
If you wouldn't let someone else organize your bedroom and decide where things go, why are you fine with letting a piece of software do the same thing to your information?
Problem: Opaque Methodology for EdTech Selection, No representation for teacher in the process
The software: gamified worksheets with a dashboard attached
Spend a few hours with a representative selection of K-12 ed-tech platforms and a pattern emerges fast. They are gamified worksheets. The interaction loop is: read a prompt, click an answer, get a green check or a red X, watch a progress bar advance, collect a streak, maybe earn a badge. The underlying content is whatever someone wrote in 2014, lightly remixed and adaptively reshuffled. The pedagogical claim is that the system "adapts to the student's level." In practice it adapts by giving easier problems when the student gets one wrong and harder problems when they get one right — which is what every halfway-attentive teacher already does in person, and is not the magic of personalization.
What these platforms are actually good at is producing data. Engagement minutes. Time-on-task. Number of attempts. Streaks. Districts love this, because districts are evaluated on metrics, and the platforms package the metrics into dashboards that look like progress. The trouble is that none of those metrics are learning. They are compliance, dressed in the language of analytics. A student who has logged two hundred minutes on a math platform this month has done two hundred minutes of clicking; the dashboard cannot distinguish that from two hundred minutes of thinking.
The deeper failure is what the platforms do to the shape of a lesson. A worksheet on paper is a discrete object — the student does it, the teacher reads it, you both move on. A platform is a perpetual feed. There is always one more problem, one more streak to maintain, one more notification. The student's relationship to the material starts to resemble the relationship they already have with every other app on every other screen: scroll, react, scroll, react. The medium is the message, and the message is consumption.
The Chromebook Problem, or "dumb hardware makes for dumber people"
The Chromebook is the cheapest possible computer. That is its actual selling point, dressed up in Google's marketing as "simplicity." For a district trying to put a device in every backpack on a public-sector budget, the math works: a $250 laptop, a free productivity suite, free email accounts, and a fleet-management console that lets one IT staffer wipe and reissue a hundred devices in an afternoon. From the procurement side it is an obvious yes.
From the student's side it is a screen for consuming web pages. You cannot install a real IDE on it. You cannot run a real compiler. You cannot drop into a terminal and see what your computer is actually doing. You cannot edit a config file, or break something and fix it, or learn the satisfying frustration of the first time you brick a working system. The hardware encodes an assumption about who the user is: a content consumer, not a tool user, not a maker. There is nothing under the hood because there is no hood.
The generation of computer-literate adults the public sector is implicitly trying to reproduce — the engineers, sysadmins, computational scientists — did not become computer-literate because school taught them to be. They became computer-literate because they were handed a machine with edges you could cut yourself on, and they were curious enough to poke around. The Chromebook has been deliberately designed to have no edges. There is no Terminal.app. There is no Program Files. There is no /etc/. Everything happens inside the browser, inside an account, inside a session that gets wiped at logout. The metaphor a Chromebook teaches is the metaphor of a kiosk: arrive, perform a transaction, leave.
This is the floor that everything else sits on. The ed-tech platforms are bad partly because they are bad and partly because the device beneath them cannot run anything else. The flat-organizational-vocabulary problem from earlier, kids who can't grasp file hierarchies, is downstream of the same fact: there is no file hierarchy visible on a Chromebook, because there is essentially no local filesystem the user is allowed to think about. The cloud is the only place files can live, the browser is the only environment they can live in, and the dashboard is the only thing measuring whether anyone is paying attention. The whole stack, from hardware up through software up through evaluation rubric, is internally consistent. It is also, when you look at what it produces, indefensible.