Closing and Layering a Term System#

Within a closed term system, communication is happening on another level: AI does not have to wait for a human to provide details of the system. After a human introduces the idea and formulates the problem or the desired state, whatever the closed term system is designed to be initialized with, a human interacts with AI via the formal texts that define the system. This effectively makes it an interface between a human and AI.

Reliability of a closed term system depends on the structure of the closure and the consistency of the levels of detail.

Structure of the Closure#

Ideally, every term in the system is defined within the system through its relations to other terms. The chain of definitions cannot go on indefinitely: at some point, there are terms the system cannot define. These are foundation terms. The system owns them, but it does not own their definitions, which rest on what the reader already knows. Instead of defining foundation terms, the system introduces them in a dedicated, formalized space, using expected terms carefully chosen for this particular system. In this way, a closed term system gives them a stable base. Without it, foundation terms invite different readings, and since every other term is built on them, a misread foundation term distorts the understanding of the whole system.

A problem statement is an example of such a space. It is written before any solution exists, and it names what the solution will be about in words the reader already understands. Everything the solution later defines is built on the terms the problem statement introduced.

Levels of Detail#

The consistency of the levels of detail deals with another problem: without it, the structure of the system becomes obscure. What does an online store consist of? A product catalog, a shopping cart, a checkout, and a message queue. Each of these terms can be well defined, and the list looks complete. Yet the message queue is how the checkout passes orders on for processing: it is not a part of the store next to the checkout but a part of how the checkout is implemented. The answer flattens two levels of detail into one.

To avoid this, each term must be placed on exactly one level. Although glossaries are the designated place for term definitions, it is more common for definitions to be included in structured documents that present information in the form of a narrative.

The level of detail can apply to any structural unit, ranging from a sentence up to the whole document, in the sense that all terms in the unit belong to the same level. If the level of detail is associated with a small unit, such as a sentence, switching from one level to another happens frequently. The switch not only requires more effort to understand the meaning but also creates an additional risk of incorrect interpretation.

A natural approach is to keep the chosen level of detail within a section and declare it in the heading. This works for each section individually. When the context grows to include more than one section, the switch becomes much harder to notice: the declaration stays with its section, while the reader’s context does not. A good example of this problem is cross-references, where the reader is recommended to continue reading elsewhere for more detail and arrives at a different level in the middle of a thought.

The most stable control of the level of detail is through whole documents whose kind implies the level of detail or which state it explicitly. Documents whose structure is predefined are particularly reliable in this respect, as each section must follow a set of requirements. Links to other documents make switching the level of detail a predictable transition or continuation: unlike a cross-reference into another section, a link leads to a document whose level the reader knows before starting to read.

Tiers: Term Catalogs in Document Types#

When a closed term system is implemented in a set of documents, the level of detail becomes its organizing principle. Suppose the level of detail is applied consistently per document, and each document is treated as an instance of a document type: a set of predefined sections whose content is created and modified according to specific requirements. Then the whole system can be reasoned about as a network of relations among those document types.

When terms are extracted from an instance, the associated document type provides simple criteria for determining which terms belong together. The terms defined in an instance of a specific document type form a closed term system, too: terms defined in instances of more general document types are known to its reader and work within it as expected terms. Provided the level of detail covers the whole document, the terms in the catalog can be evaluated as expected or provided terms, which reveals problems such as missing term relations or excessive details.

The catalog of terms defined in one instance constitutes a separate tier within the closed term system being designed. If each term is defined in only one instance, it belongs to exactly one tier, and the document type of that instance gives it its level of detail.

Orders: Arrangements of Tiers#

With an effective arrangement of tiers, a closed term system becomes repeatable: the same arrangement can be used to build new systems. An order is a template for such an arrangement. It is a lightweight construct. It mandates document types, which contain no content of their own, associates each of them with a level of detail, and prescribes which more general document types supply terms to which more specific ones.

When an order is instantiated, the instances of its document types form a closed term system as the product, and their term catalogs become its tiers.

+ References#

+ Page Info#

Version	1.0
Revision	2026-10-16
Written.by	Borys Belinsky (borys.belinsky@protonmail.com)
Assisted.by	Anthropic Claude, Opus 5.5