About

The Encyclopedia of Abstractions is an exploratory, long-duration intellectual project investigating cross-domain abstractions, solution archetypes, and structured reasoning in collaboration with generative AI systems. It is not a finished ontology, a settled taxonomy, or an authoritative reference work. It is best understood as an ongoing experiment — a substrate for thought and experimentation.

Contact

If you have any questions or comments, feel free to contact me via email, Facebook, or LinkedIn.

· <Mouse over for email address…>  
· Facebook  
· LinkedIn

I'm Kurt Zoglmann, and this is a project I work on in my own time.

I would describe myself as a systems thinker. With a B.S. in Information Systems from Kansas State University and a career in software development, I enjoy abstractions and pushing at the boundaries of what's known. Building software is rarely just a matter of using the abstractions handed to you by libraries and frameworks. The interesting part — the part that feels like craft — is defining new ones: deciding where the seams of a system should fall, what to name, what to hide, what to expose, so that a rat's nest of detail collapses into something a person can hold in their head and reason about.

This project is intellectual play at scale — the most ambitious playground I work on in my off-hours. I mention this not to be self-indulgent but because it explains the project's character: it is driven by genuine curiosity rather than by an institutional mandate or a commercial plan, and that shapes both its ambition and its limits.

For a small worked example of the kind of cross-domain abstraction this catalog is about, here is one I wrote and never published: The Nothingness Ladder — a hierarchy of what "nothing" could mean, from the quantum vacuum down to logical non-existence. The interesting move is not picking which rung is right; it is that building the ladder dissolves much of the apparent disagreement between physicists and philosophers about "what is nothing" by showing they have been talking about different rungs. The same instinct drives this whole project.

As an interesting tidbit: I am also the creator of Morse Code Ninja, a free resource for learning Morse code that I have built and maintained over many years (you'll find me on the air as AD0WE). On its surface that is a different world — amateur radio, dits and dahs — but to me the two projects come from the same place. Morse code is encoding made tactile: meaning carried on a single on/off channel, information theory you can hear. And like this one, Morse Code Ninja is a large, structured, patiently-built thing made in my own time and given away in the hope that someone finds it useful.

The same temperament runs through both — a love of structure and abstraction, and a habit of turning something vast into something elegant a person can actually learn. If you found your way here from the ham radio world: welcome, and 73.

What it is

At its core, the Encyclopedia documents prime abstractions — cross-domain patterns that are at once broadly universal and fundamental to how we reason and solve problems. A prime abstraction is not merely a label or a category; it is a principle that recurs across at least three distinct domains of human knowledge — feedback, equilibrium, hierarchy, sovereignty, recursion, and several hundred more — and that earns its place by helping us manage complexity, surface the relationships beneath a system, and carry knowledge from one field into another. They are among the basic building blocks we use to construct models of the world, and each entry tries to state the pattern, its components and relationships, and where it does and does not apply — so that it can be reasoned with, not merely named.

Alongside the primes sit solution archetypes — recurring shapes of intervention: the moves that show up again and again when people try to solve structurally similar problems, regardless of the domain they appear in. If a prime names a pattern in the world, an archetype names a pattern in how we act on it. Around both sit the supporting apparatus — the components and mechanisms the archetypes are built from, the relationships among primes, and a set of conceptual essays about how the whole thing is meant to behave as an instrument of cross-domain reasoning.

Not every useful abstraction is universal. Many are domain-specific — specialized to a single field or a small cluster of related ones, like object-relational mapping in software engineering or acid–base equilibrium in chemistry. These are not the core focus, but they matter: some are simply valuable within their field, and some may prove to be bridges to other domains once the connection is recognized. The long-term aim is to gather them into discipline-specific Domain Packs — collections grouped by field for the experts who work there, and opened to everyone else in case their cross-domain potential is waiting to be discovered. Domain Packs are meant as extensions of the prime set, never substitutes for it.

At this time, the catalog holds 655 prime abstractions and 625 solution archetypes, and runs to something on the order of 6.3 million words. That number is worth pausing on, because it is central to everything else on this page.

Why it could only exist now

I want to be direct about the role of artificial intelligence in this work, because it is not incidental — it is constitutive.

This project would have been impossible before large language models. I don't mean it would have been hard, or expensive, or slow. I mean it could not have been done. No single human being — not the most widely read scholar alive, not the most brilliant generalist — could personally identify the prime abstractions across mathematics, physics, biology, law, economics, linguistics, anthropology, and the rest of the domains of human knowledge, and then identify the solution archetypes that cut across those same domains, and then synthesize all of it into a few million words of coherent prose. The breadth required exceeds any individual's reach, and the volume of writing exceeds any individual's lifetime.

What changed is that a new mode of work has become possible — one I suspect will become increasingly common. Large language models can serve as cognitive scaffolding: they hold a vast, cross-disciplinary range, and they can draft and synthesize at a scale no person can match. When prompted well, they can challenge one's ideas and preconceived notions, and surface hidden assumptions and biases. And used well they enable a single motivated person to operate at a level of intellectual productivity that would otherwise be superhuman. The Encyclopedia is, among other things, an artifact of that mode — a demonstration of what one person plus capable AI systems can attempt when the goal is genuinely vast.

That is also why I think the project is interesting beyond its content. It is a worked example of human–AI collaboration aimed not at a narrow task but at an open-ended, knowledge-spanning intellectual endeavor.

How it was made, and what to trust

Because the AI's role is constitutive, I owe you an honest account of the project's epistemic status. Here is the uncomfortable truth, stated up front: a corpus of this size cannot be fully validated by one person, and I am one person.

The catalog is too large for any single human to read closely, let alone edit line by line, let alone verify against the primary literature of a dozen disciplines. Ideally, a work like this would be reviewed by domain experts across every field it touches — a mathematician checking the mathematical primes, a legal scholar checking the legal ones, and so on. That kind of validation is exactly what one person cannot provide across all of human knowledge. So I won't pretend the corpus has it.

What I have done is take the problem of error seriously and work at it at scale. Considerable effort has gone into reducing and detecting hallucinations — the confident fabrications that language models are prone to. For the references in particular, work has been done in bulk to verify that the cited sources actually exist, and to identify the specific claims the model was less certain about, or that seemed to call for support from an authoritative reference, and to anchor those claims accordingly. That has been the approach throughout: not a guarantee of correctness, but a systematic, repeatable effort to raise the floor.

It is not perfect. I would, in fact, assume there are errors in here — misattributions, overconfident claims, the occasional citation that doesn't say quite what it's invoked to say. If you find one, I'd genuinely like to know. But my considered view is that the project is good enough to stand on its own as something useful, and that its usefulness does not depend on its being flawless. A map can be wrong in places and still be worth having.

So: read it as a thoughtful, large-scale, AI-assisted synthesis offered for exploration — not as a peer-reviewed reference you should cite without checking. Where it points you somewhere interesting, follow up at the source.

One scope note. The site you are reading is the public face of a larger working repository. Some internal artifacts — provenance ledgers, drafting workspace, ad-hoc experiment scripts, project logs — are referenced by the conceptual papers as evidence that a particular piece of work was done, but are not published. If something doesn't serve a public reader's purpose, I generally don't publish it; the artifacts that do serve a public purpose (raw experimental data, scoring scripts, downloadable datasets) are on the Downloads page.

Who it's for

The project has three audiences in mind, and it tries to serve all three.

The first is the cross-domain thinker — anyone who reasons across boundaries and wants a structured place to do it. If you've ever sensed that a problem in your field rhymes with a problem in someone else's, the catalog is built to make that rhyme explicit: to let you find the shared structure beneath two superficially different situations, and to see the menu of moves that have worked on that structure elsewhere. It is meant to be a substrate for thought.

The second is educators and students. I hope the Encyclopedia can become a foundation for teaching abstract thinking directly — helping students learn to recognize the abstractions operating in a problem, and then, increasingly in partnership with AI, to build models they can reason over: both domain-specific models and the stripped-down meta-models that let an insight travel between fields. Abstract reasoning is rarely taught as a skill in its own right; a structured catalog of the recurring patterns, with worked cross-domain examples, is one way to make it teachable.

The third is AI and reasoning researchers. How machines (and people) reason across domains — how an abstraction is recognized in an unfamiliar situation, retrieved from memory, and mapped onto a new problem — is an open and genuinely hard problem, and the Encyclopedia is, deliberately, a large structured testbed for it: a corpus over which one can study how cross-domain abstractions behave and how they are found. (Several of the conceptual essays here grew directly out of experiments in that spirit.) Researchers are also the audience most likely to use the corpus rather than read it — as a substrate for augmented abstract reasoning, a source of synthetic training data, or seed material for generators and other tools built on its ideas. That is a first-class use, arguably the one with the most leverage, and part of why the catalog is structured the way it is.

An invitation, and the road ahead

I've thought a fair amount about what happens when something like this is put into the world, and I see two honest possibilities.

The first is that it is largely ignored. That is the most likely outcome, and I'm okay with that. The project has already paid for itself in the making; it is the kind of thing I would have built whether or not anyone ever read it.

The second is that the right people encounter it, recognize what it is, and want to do something with it. If that happens, the natural next step is community — because the thing this project most needs is exactly the thing one person cannot supply: validation, correction, and extension by experts across the many fields it spans, in the spirit of a wiki. Others would build on it rather than within it — researchers and tool-makers — and educators would adapt it for the classroom. A catalog of cross-domain abstractions is, almost by definition, a collaborative object. If the project resonates with you in any of these ways — to correct an entry in your area of expertise, extend the corpus, build something on top of it, bring it into your teaching, or just argue with its premises — I'd be glad to hear from you.

Either way, the work will never be finished, and I mean that as a description rather than an apology. Human knowledge keeps growing, and I expect generative AI to accelerate that growth rather than slow it. A catalog of humanity's recurring patterns of thought is therefore a moving target — something to be tended and expanded indefinitely, not completed and closed. That open-endedness is part of what makes it an ongoing experiment, and part of why I find it worth doing.

Thank you for reading, and for spending a little of your scarce attention on it.

— Kurt Zoglmann