The Arc
What does "computer" mean?
Eight times in eighty years, the answer changed. Each shift was controversial in its moment and invisible within a decade. The current shift — "thing that synthesizes from intent" — is the ninth.
Part 01 · The arc
Each era added a layer of abstraction between the user and the silicon.
Eight transitions, each one controversial in its moment, each one invisible within a decade. The pattern: a new abstraction enters, the old definition of "computer" becomes obsolete, the abstraction becomes substrate, and the next abstraction is built on top.
Each colored band is an abstraction layer that became substrate for the next. The user no longer has to think about silicon. Soon they will no longer have to think about specification. The trend is monotonic: the human writes less, the machine infers more.
Part 02 · The eras
Eight redefinitions of one word.
For each boundary, the question is not "what was invented" but what changed about what people meant when they said "computer." The invention is downstream of the definitional shift, not upstream of it.
Part 03 · The stack
Each abstraction became invisible substrate for the next.
In its own era, every abstraction was attacked: real programmers don't need an operating system. Real users don't need a mouse. Real computer owners don't need to be on a network. Each became invisible within a decade.
Each layer remains under the next. None has been removed. The user of 2026 is still ultimately using silicon — just through eight intermediate layers, each of which made the layer below it invisible.
Part 04 · Where we are now
The mystique is the transition.
Every previous era boundary produced the same confusion. The word "computer" temporarily meant two incompatible things at once. The pre-PC era ended when the word stopped meaning "thing in a server room" and started meaning "thing on a desk." The pre-mobile era ended when it stopped meaning "thing on a desk" and started meaning "thing in your pocket."
We are in that fog right now. "Computer" is shifting from "thing you give instructions to" to "thing that synthesizes from intent." Both meanings are operating simultaneously. The confusion shows up as three questions that are systematically blurred together:
What can it do?
A bounded question, answered by enumeration. The same question one would ask of a database, a compiler, or a calculator.
How well does it do it?
A measurement question, answered with numbers, against alternatives, on real workloads. Falsifiable.
Why should it do it?
A comparison question. Against what other tool? At what cost? With what risk? Engineering, not vibes.
The current era collapses the three into a single demo: "look, it wrote a poem." Capability is shown, performance is assumed, justification is waved at. Every previous era boundary did the same thing, until the abstraction stabilized and the questions could be asked separately again.
The mystique is not a property of AI. It is the temporary fog of a category in transition.
What survives the fog is what survived every previous transition: tools that measure honestly, deliver work at the right cost, and let the user understand what they are paying for. The abstraction is going to stabilize. When it does, AI will be substrate — invisible, taken for granted, available to the next layer above it. The era after this one is already implied by it.
In the meantime, the work of this era is to measure what the abstraction actually costs and to engineer tools that route the work to the right gear. That is what Synthesis is for.
Continue
Read the map.
The thesis names the moves. The axes give the coordinates. The instrument shows it running.