The Unit of Knowledge Is Not the Page. It Is the Change.
Pages describe how things are. To understand how the world evolves, we need to record what changed, when it happened, and which evidence supports it.
The Internet is very good at answering one question: what is this like now?
A company page describes what the business does. A biography summarizes a person’s career. A reference page explains a law, a technology, or an organization. When the page is maintained well, it offers a reasonable snapshot of the present.
But many of the questions that matter most are different:
How did we get here?
The current state is not enough to answer them. We need to know what changed, when it happened, what caused it, and which evidence allows us to trust it. We need a sequence, not only a description.
That distinction may seem small, but it affects nearly everything we try to understand online. A page tells us where we are. Changes explain the path.
An updated page can also erase information
When a page is updated, we usually assume it has improved. The old fact disappears and the new one takes its place. The document becomes more accurate, cleaner, and more useful to anyone who only needs the present.
Yet something important can disappear during that update: the transition.
If a company changes leadership, an organization shifts its priorities, or a policy enters a new phase, replacing the old state with the new one preserves the result while potentially erasing the story. We know what is true today, but may no longer see clearly what stopped being true yesterday.
The page is not wrong. The problem is that an updated photograph does not necessarily contain a film.
Web archives can help us retrieve previous versions, but they require readers to compare entire documents and discover the changes themselves. The history exists as a sequence of copies. It rarely exists as explicit knowledge.
Facts matter; transitions explain
An isolated fact can be entirely correct and still be insufficient.
Knowing who leads an organization is useful. Knowing that this person replaced someone else after a particular decision explains much more. Reading the current version of a policy is necessary. Understanding which provision changed, when it changed, and what followed makes its evolution intelligible.
Time turns facts into a structure:
- something was true
- an event occurred
- a condition changed
- a new state emerged
- a source recorded the transition
The basic unit is no longer only the fact “A is B.” It is also the change “A moved from B to C.”
The second format is closer to how we build explanations. We do not merely accumulate statements; we look for causes, consequences, decisions, and turning points. We want to understand what persists and what transforms.
Change is a more demanding unit
Recording changes sounds simple until we try to do it rigorously. New information is not always a change, and a new headline does not always describe a new event.
Three publications may report the same announcement at different times. A statement may repeat an existing intention. A number may be corrected without changing the broader trend. A rumor may look like a transition until a primary source disproves it.
Representing a change therefore requires several questions:
- which entity changed?
- what was its previous state?
- what is its new state?
- when did the transition happen?
- which source supports it?
- is it genuinely new or a repetition?
- is the information still current?
A page can display the latest value. A temporal knowledge system must also preserve the reasoning that connects that value to the one before it.
This adds complexity, but it creates something a collection of documents cannot provide on its own: traceability.
Sources are not footnotes
When knowledge is organized around changes, provenance stops being a decorative detail. It becomes part of the change itself.
It is not enough to state that something happened. We should be able to return to the document, announcement, record, or reporting that allowed us to observe it. We should also distinguish between the source that announces a decision, the source that verifies its implementation, and the source that analyzes its consequences.
One transition may require several perspectives. A primary source provides precision about what was declared. Independent reporting may reveal what happened in practice. A later update may correct the initial interpretation.
Evidence does not weaken the narrative. It makes the narrative revisable.
That capacity for revision is essential because living knowledge is not about being right once and freezing the answer. It is about being able to correct the answer without losing the memory of why it changed.
From organizing documents to organizing time
Search engines organize documents by their relevance to a query. Feeds order them by publication date. Encyclopedias assemble a relatively stable synthesis. Each format solves a different part of the problem.
An evolving story needs another layer: a way to organize time around the thing that persists.
The entity—a person, company, institution, place, or technology—provides the thread. Changes form the sequence. Sources support each step. The current state then appears as the provisional consequence of a verifiable history.
This structure makes more useful questions possible:
- what changed since my last visit?
- which decision produced the current state?
- which parts of the story remain open?
- which claims have been corrected?
- where do sources agree or disagree?
Traditional search tries to find the right document. A memory organized around change tries to reconstruct the right evolution.
It also changes how we summarize
A conventional summary compresses a set of texts. It may be excellent and still combine information from different moments as though everything were simultaneously true.
A temporal summary has another responsibility. It must recognize which information was valid before, which is valid now, and which event separates the two states. It does not only reduce words; it preserves the direction of change.
This matters especially for artificial intelligence. A model can write a convincing answer from incompatible fragments because fluency does not guarantee temporal understanding. Given facts without dates, states without transitions, or sources without provenance, it can produce an explanation that is grammatically flawless and chronologically impossible.
Giving an AI system better context does not simply mean adding more documents. It means providing a structure that distinguishes past, present, uncertainty, and correction.
The present is only the latest version
We tend to treat the present as a conclusion, even though it is almost always provisional. Companies will change again. Technologies will find new uses. Policies will be amended. Many interpretations that appear solid today will eventually need revision.
A traditional page tries to stay current. A page with memory also tries to explain every update.
It is the difference between replacing and accumulating; between erasing an old state and turning it into context; between displaying the answer and preserving the reasoning that brought us to it.
If we want the Internet to become more than an archive of documents, we will need systems capable of representing movement. Systems in which a correction does not destroy the past, a novelty is not confused with a repetition, and a claim can be traced all the way back to its source.
Pages will remain necessary. Documents will remain the raw material. But perhaps the most useful unit for building knowledge is neither of them.
Perhaps it is the change.
A page can tell us what the world is like. Should it not also help us understand how it became that way?