We are taught to think of abstraction as a simple convenience, a kind of mental tidying-up. To abstract, in this conventional view, is to hide messy details behind a clean interface, allowing us to manage complexity without getting bogged down in the machinery. But what if this pedagogical story obscures a much deeper and more powerful truth?
In a sprawling body of work, the philosopher and computer scientist Flyxion presents a profoundly different understanding of abstraction—one that recasts it as a fundamental process shaping computation, ethics, and even reality itself. This article distills the four most impactful and counter-intuitive ideas from that work, each of which promises to change how you see the world.
Remember learning algebra? Amid the variables and equations, you were given a simple, almost trivial, rule for doing arithmetic: BEDMAS / PEMDAS. The first and most important step was always to evaluate what’s inside the parentheses first. It seemed like a mere convention for getting the right answer.
But according to Flyxion, this elementary rule is structurally identical to the innermost evaluation principle that governs the most advanced computational systems in existence—from the lambda calculus, the foundational language of computation, to the novel Spherepop Calculus.
The principle is the same everywhere because both systems are fundamentally about resolving nested dependencies before dealing with the outer context. The rule is to resolve the deepest, most nested obligations before you can reason about the larger structure that depends upon them.
What begins in school as a directive for “doing arithmetic in the proper order” can be understood as an informal apprenticeship in the deep logic of all computation:
- Identify a local cluster of dependencies
- Collapse that substructure into a simpler form
- Iterate outward until no further reductions remain
This reframes computation not as a specialized technical skill, but as a fundamental pattern of structured thought we all learn. The act of solving an equation on paper is, in its deep structure, the same act unfolding at the heart of our most complex technologies—revealing an astonishing unity between everyday reasoning and formal systems.
Public discourse on AI risk remains captured by a mythology of malevolent superintelligence. Flyxion’s work—particularly the essay The Autonomy of Refusal—argues that this focus is a dangerous distraction.
The real peril lies not in what a system might someday want to do, but in what today’s systems cannot do: stop.
At the core of this argument is the concept of refusal, redefined not as mere negation but as a positive and uniquely human capacity. Refusal is the non-instrumental ability to suspend execution—to withdraw from a procedure without offering an alternative.
The canonical literary example is Melville’s Bartleby the Scrivener, whose quiet declaration—
“I would prefer not to.”
—is a form of pure refusal. It is not negotiation or protest; it is a withdrawal from the logic of participation itself.
Scalable systems—from ancient bureaucracies to modern large language models—achieve their power precisely by amputating this capacity. To scale, operations must be predictable and repeatable. Contextual judgment must be stripped away.
Scalable intelligence resides in the absence of refusal.
As a consequence, responsibility, alignment, and safety cannot be internal properties of such systems.
One devastating real-world consequence is the collapse of apprenticeship. Institutions historically hired juniors not for immediate productivity, but to absorb uncertainty, encounter edge cases, and slowly acquire judgment under supervision. This slow, interruptible layer is where refusal is learned.
When non-refusing abstractions outperform humans economically, apprenticeship becomes irrational. Institutions continue to depend on scarce senior judgment while systematically destroying the pathways that create it—optimizing themselves into long-term fragility.
The risk, then, is not future hostility but present inexorability.
Abstraction is necessary for thought—but it is never innocent. Every abstraction has an ontological cost: it discards detail. When that detail involves lived reality, abstraction becomes a weapon of disappearance.
The danger arises when abstraction reduces a complex reality to a disposable representation that serves a narrow system logic:
- A human becomes a datapoint
- A worker’s body becomes a unit of labor
- A forest becomes a carbon metric
In each case, abstraction makes the world more tractable by making parts of it invisible—and therefore disposable.
The joke that
“many crimes have been committed in the name of abstraction”
is not merely humorous; it is philosophically exact.
Abstraction can function as encoded indifference. When systems reduce our cognitive load by simplifying the world, they often reduce our ethical load as well. This demands a deeper responsibility: we must ask not only what our abstractions allow us to do, but what they force us to forget.
The final and most foundational idea overturns the conventional definition of abstraction itself.
We are taught that abstraction hides messy processes behind clean interfaces. Flyxion argues this is backward.
In this model, abstraction is not a wrapper around unresolved work—it is the result of work that has been successfully completed. A clean, stable interface becomes available only after the internal complexity it depends on has fully resolved.
You don’t get a function’s simple input-output behavior by ignoring its internal chaos. You get it because that chaos has finished executing.
As in software, so in hardware: in asynchronous circuits, a clean logical value becomes available only after a signal settles into a determinate state, concluding its local work.
Abstraction is not the negation of detail; it is the successful execution of detail.
This reframing has profound implications. Robust systems are not built by hiding unresolved problems, but by ensuring that each component has truly completed its task—becoming stable and composable.
The clean surfaces we rely on are not illusions. They are achievements.
Taken together, these four ideas offer a radically new lens:
- Elementary arithmetic reveals the deep logic of computation
- The greatest technological risk is inexorable obedience, not rebellion
- Simplifying concepts carry real ethical weight
- Stable abstractions are monuments to resolved complexity
This understanding of abstraction is more than an intellectual curiosity. It poses a fundamental question:
If abstraction is the engine by which our world is built, what does it mean to build it well?
Flyxion’s work offers a demanding answer: an ethical abstraction must not erase the moral, ecological, or phenomenological substrates that give rise to it.
We must know precisely what we are choosing to forget.