Personal Computing
Computing whose unit of analysis is one person and the work she is actually trying to do — the older sense of the phrase, before it became a retail category, and what would have to change for it to mean that again.
The phrase personal computer has had two lives. In the first, beginning roughly with Engelbart’s 1968 demonstration and codified by Kay and Goldberg’s Personal Dynamic Media in 1977, the adjective personal was load-bearing: it picked out a machine whose primary virtue was that it amplified the thinking of the individual sitting in front of it. The Dynabook was not a small mainframe, and the Alto was not a thin client. They were proposals for a kind of medium one could think with, comparable in ambition to writing or to mathematical notation.
In the second life of the phrase, personal became almost vestigial. A personal computer, today, is a category in a retail store. The objects in that category are mostly portals — surfaces through which their users reach software, data, and decisions that live elsewhere. The work is not done on the machine; the machine is where the work is displayed. The computer is personal in the way a hotel room is personal: temporarily yours, almost entirely furnished by someone else.
I want a name for what got lost between those two lives, and I want to spend time on what it would take to recover the parts worth recovering.
The original ambition
Engelbart, Kay, and Nelson did not agree on every particular, but they shared one premise: a computer’s job was to expand the user’s capacity for serious intellectual work. Augmenting Human Intellect (1962) framed the computer as a partner in sense-making — note-taking, outlining, linking, collaborating, modelling. Kay’s A Personal Computer for Children of All Ages (1972) framed it as a medium of authorship, on which the user was expected to read and to write, in the strong sense of that word. Nelson’s Computer Lib / Dream Machines (1974) framed it as a tool for liberation from bureaucratic information systems.
Across the three, the assumption was that the user was a maker. The machine’s chief affordances were creation, composition, and connection, and the ratio of authorship to consumption on a well-designed personal computer was supposed to be high.
What changed
The path from there to here is well known, but worth saying once with conviction. The personal computer, in becoming successful, became a mass-market appliance, and mass markets are made of users who, by definition, do not all want the same things; the easiest way to serve a heterogeneous mass is to make the device do less by default and offer to do more by purchase. Authorship became a niche; consumption became the path of least resistance. The economics of the app store — predictable revenue from rentable software — accelerated the trend, and the cloud, by relocating canonical state off the device, finished it. The machine that was supposed to be the user’s medium became, in many cases, a viewing window onto someone else’s.
This is not a complaint about any particular company. It is the predictable result of optimising a general-purpose medium for the metrics of consumer electronics.A medium optimised for engagement is, almost by construction, optimised against reflection. Bret Victor has been making a related argument, in different language, for over a decade.
Personal computing as a research stance
I find it more useful to treat personal computing as a research stance than as a product category. The stance is roughly this: the unit of analysis is a single person and the work she is actually trying to do, and the question is what software would look like if it were designed in service of that.
Several commitments fall out of taking this seriously. The first is locality — her data lives somewhere she can find, inspect, and back up, not somewhere she rents from a vendor that knows her by an account ID. Durability follows from it: the artefacts she makes outlive the applications she made them in, and a tool’s disappearance does not take her work with it. Comprehensibility is the quieter sibling of these two, and the one most often skipped — the system represents her work in a way she could, in principle, understand, even if she chooses never to look. There is also authorship, which is what malleability looks like from the user’s side: the system reserves room for her to bend it, not just to configure it within whatever envelope the vendor has approved this quarter. And continuity, which sounds modest until you notice how much modern software is at war with it: her tools are the same on Tuesday as they were on Monday, unless she chose to change them.
None of these are exotic. All of them are, in 2026, unusual.
Open questions
The questions that hold my attention sit at the intersection of design, infrastructure, and economics, and they are mostly about how to give an ordinary person more leverage over her own computing without asking her to become a sysadmin.
The first is structural: how should personal data be organised so that she can move between applications the way she moves between pens — choosing the tool for the task, not the platform for the lifetime? The second is atmospheric, and harder to phrase well. What does a home feel like, in software, as distinct from a feed? Most of what an ordinary user encounters on her devices is arranged for her by something else; I would like to know what it would mean for the arrangement to be hers. The third is economic, and I have less patience for it because the answer is not really a research question: I want tools whose value to the user accrues over years, rather than tools whose value to the vendor accrues over years, and I want to understand what kinds of design choices and what kinds of business models make that possible. There is also a structural puzzle about the operating system itself, which on most devices has been quietly hollowed out into a thin layer beneath the browser and the app store; what an OS is now for, in any deep sense, is no longer obvious to me. And, more diffuse but perhaps the most important: whether we can build personal computing environments that invite quiet, long, focused work without that being a luxury feature of a productivity-influencer subscription.
Underneath these is a longer question I do not yet know how to phrase, about whether personal computing is even possible at the scale at which software now operates. Most of the systems my computing depends on are not, in any meaningful sense, mine. They are services I rent, and they work because millions of other people rent them too. Recovering the older sense of personal may require accepting that some part of one’s computing should be deliberately small, deliberately local, and deliberately independent — even at a cost in convenience.
Pole stars
The reading list for this area is mostly old, and I am not embarrassed by that. Augmenting Human Intellect; Personal Dynamic Media; Computer Lib / Dream Machines; Bret Victor’s essays, especially A Brief Rant on the Future of Interaction Design. Among contemporary work, the Local-First Software essays from Ink & Switch and the steady output of the Future of Coding community are the most serious continuations of the line.
A working stance
The aspiration, in the end, is modest: to treat the computer in front of a person as a place she can actually live in, rather than a kiosk through which she negotiates with services that do not know her. That used to be the default. It can be again — not by going backwards, but by remembering what the adjective in personal computing was originally doing, and designing as if it still meant it.