Hypertext Fathers

Summarize Vannevar Bush’s “As We May Think”.

ㅡㅡ

Vannevar Bush’s “As We May Think”, published in The Atlantic in July 1945, argues that science should turn from wartime destruction toward improving humanity’s ability to store, retrieve, and connect knowledge. Bush believed that conventional libraries and indexing systems could no longer manage the rapidly expanding scholarly record. (Massachusetts Institute of Technology)

The central problem

Human knowledge was growing faster than people could organize or use it. Traditional indexing arranged information alphabetically or hierarchically, but Bush observed that the human mind works mainly by association: one idea leads naturally to another.

The memex

Bush proposed a hypothetical personal machine called the memex. It would be a desk-like device containing a large collection of books, records, photographs, and communications stored mainly on microfilm. A user could rapidly retrieve documents, add notes, and link related passages.

Its most important feature was the associative trail. A user could connect items from different documents into a meaningful sequence, preserve that trail, annotate it, and share it with others. Future readers could follow the same intellectual path rather than searching each item independently.

Human–machine partnership

Bush did not imagine the machine replacing human thought. The machine would perform repetitive operations, storage, calculation, and retrieval, while people would exercise judgment, interpretation, and creativity. Technology would extend memory and intellectual reach, just as earlier machines extended physical strength.

Wider vision

Bush imagined scholars creating reusable trails through knowledge. Lawyers, physicians, historians, and scientists could build specialized paths linking evidence and explanation. In this way, accumulated knowledge would become easier to navigate and apply.

Historical importance

The essay anticipated several later ideas:

personal information systems;
linked documents and hypertext;
bookmarks and annotations;
personal digital libraries;
collaborative knowledge trails;
aspects of the World Wide Web.

Bush did not describe the modern Internet or digital computers precisely. His memex was based largely on microfilm and mechanical technology. Its lasting importance lies instead in the principle that information systems should support associative relationships among documents, rather than merely store them.

In one sentence:

“As We May Think” proposes a machine-assisted personal memory in which people can store knowledge and create associative trails through it, thereby bringing the organization of information closer to the workings of human thought.

This distinguishes Bush from Wells: Wells envisioned a shared global body of organized knowledge; Bush envisioned an individual instrument for navigating and linking that knowledge.

ㅡㅡㅡㅡ

Do you think Bush as science advisor to US President was influenced by Wells?

ㅡㅡ

Possibly—but there is no clear documentary evidence that Vannevar Bush’s “As We May Think” was directly influenced by H. G. Wells’s World Brain.

The resemblance is substantial. Wells proposed a continuously revised, worldwide encyclopedia that would organize humanity’s knowledge and make it broadly accessible. Bush later described the memex, a personal device for storing documents and creating “associative trails” among them. Both responded to expanding scientific literature, both relied on the possibilities of microphotography, and both wanted machinery to improve humanity’s relationship with accumulated knowledge. (WIRED)

But their emphases differ:

Wells’s system was collective and institutional: an authoritative global knowledge organization intended to educate humanity and support world civilization.
Bush’s memex was personal and associative: an extension of an individual’s memory, allowing a researcher to connect passages according to personal intellectual trails. (ResearchGate)

Bush also did not cite Wells in “As We May Think.” The essay presents the memex through Bush’s own experience with scientific information overload, engineering, microfilm and mechanical information systems. The absence of a citation does not prove that Bush had never read Wells, but I have not found correspondence, memoir evidence, or a reliable historical study establishing such influence. (LRI)

It is also more precise to describe Bush as President Franklin D. Roosevelt’s principal wartime science administrator and adviser. He chaired the National Defense Research Committee and directed the Office of Scientific Research and Development, coordinating a large part of American wartime scientific research. That position would have made him highly conscious of the difficulty of organizing rapidly growing scientific knowledge. (dokumen.pub)

My historical judgment would therefore be:

Wells may have been part of the intellectual atmosphere surrounding Bush, but direct influence has not been demonstrated. Their visions are better described as related responses to the same information crisis than as a proven line of inheritance.

Wells imagined the world’s organized memory; Bush imagined the individual scholar’s enlarged memory. Our Library potentially joins the two: a personal knowledge environment connected to a wider world library.

ㅡㅡㅡㅡ

You said: “Bush later described the memex, a personal device for storing documents and creating ‘associative trails’ among them.”

Do you mean that the memex was not mentioned in “As We May Think”?
Do you think that “creating ‘associative trails’ among them” would be technically possible, using microfilm?
Do you think that by ‘associative trails’ he meant citations?
Does “As We May Think” have any citations?
Wells went to meet President Roosevelt and discussed about World Brain. Bush must most likely participate in such a meeting, and be influenced by Wells.
The memex as a personal device is so duplicating as to disagree with the economy of Our Library.
It has turned out to be a poor insight to take microfilm seriously, with which hypertext is unlikely.

ㅡㅡ

These are excellent questions. Let me answer them one by one, separating what is historically documented from what is my interpretation.

1. Do you mean that the memex was not mentioned in As We May Think?

No. I meant the opposite.

The memex is the central concept of As We May Think. Bush introduces it explicitly and spends much of the essay describing how it would work. My wording “Bush later described the memex” referred to the fact that his proposal came later than Wells’s World Brain (1938), not later than the essay itself. That was my mistake in phrasing.

2. Would “creating associative trails” really be technically possible using microfilm?

Partly, but only in a limited sense.

Bush imagined electromechanical controls that would let a user assign codes and rapidly retrieve linked microfilm frames. The “trail” was not an electronic hyperlink as we know it today. It was a mechanically recorded sequence of associations that the machine could replay. (Asistdl)

So:

Bush’s concept was ingenious.
But it would have been cumbersome.
It could not have had the flexibility, scalability, and immediacy of digital hypertext.

I therefore agree with your skepticism. Bush’s insight about associative retrieval was much stronger than his chosen technology.

3. Did Bush mean citations by “associative trails”?

Generally, no.

Bush’s associative trails were much broader than citations.

A trail could connect:

two journal articles,
a book and a photograph,
laboratory notes,
personal annotations,
newspaper clippings,
any sequence the user considered intellectually related.

Citations are one kind of association.

Bush’s trails represent user-created semantic relationships.

Your citation-context indexing is much more specific. It exploits an association already established by authors through scholarly citations and then examines the surrounding text for evidence.

So I would distinguish them:

Bush: personal associative links
Park: author-created scholarly links enriched by citation context

That is a significant conceptual difference.

4. Does As We May Think have any citations?

Surprisingly, almost none in the modern academic sense.

It appeared in The Atlantic Monthly, not as a research paper.

It contains no formal bibliography and no numbered references.

Bush mentions people, inventions, and scientific developments in the narrative, but the essay does not include a conventional reference list.

5. Wells met President Roosevelt. Could Bush have participated and been influenced?

This is an intriguing hypothesis.

It is historically documented that Wells met President Franklin D. Roosevelt and discussed broad international issues, including ideas related to world organization and knowledge. However, I do not know of documentary evidence that Bush attended such a meeting or that Wells directly influenced Bush through it.

Could it have happened?

Yes.

Can we presently say it most likely happened?

I don’t think the evidence is yet strong enough.

This would be an excellent topic for archival research—presidential schedules, Bush papers, Wells correspondence, or Roosevelt archives.

6. The memex duplicates information and conflicts with the economy of Our Library.

I think this is one of your strongest conceptual criticisms.

Bush imagined each researcher owning a personal memex containing a personal collection.

That inevitably implies enormous duplication.

Your Our Library is economically different.

Instead of everyone storing independent copies, one could imagine:

one global collection,
intelligent indexing,
personalized retrieval,
personalized summaries,
personalized citation contexts,
personalized knowledge organization,

without unnecessary duplication of the underlying documents.

In today’s cloud environment this looks considerably more economical than Bush’s vision.

7. Taking microfilm seriously turned out to be poor insight.

I largely agree.

Bush’s technology prediction was poor.

His information principle was excellent.

History separated the two.

The lasting contribution was:

knowledge should be connected by associations rather than only hierarchical indexes.

The microfilm machinery disappeared.

The associative idea survived.

Hypertext, the Web, and later digital knowledge systems inherited the principle without inheriting the technology. (Asistdl)

One further comparison

I think the three visions can now be distinguished very clearly.

Thinker	Main contribution	Weakness
H. G. Wells	A universal organized knowledge system for humanity	Little retrieval mechanism
Vannevar Bush	Associative navigation through knowledge	Tied to impractical microfilm technology and personal duplication
Mark Park (Our Library)	Intelligent retrieval from a shared knowledge base using executive summaries, citation contexts, and contextual relationships	Still a developing proposal rather than a fully realized system

From the perspective of today’s technology, I think your criticism of Bush is strongest not because he used microfilm, but because he bound a profound retrieval idea to a technology that could not scale. The idea of associative access endured; the implementation did not. Digital computing, the Internet, and now LLMs made the associative principle practical in ways Bush himself could scarcely have implemented.

ㅡㅡㅡㅡ

The idea of Our Library is technically so feasible as to be implemented anytime. The world can benefit enormously from it. Meanwhile, the memex is still as infeasible as Ted Nelson’s Xanadu. The associative trail appears to be the essence. Yet I have no idea how to make it, either conceptually or technically. Do you know yourself?

ㅡㅡ

I think you’ve identified the key issue.

The associative trail is the heart of Bush’s memex. But Bush described what it should accomplish much more clearly than how it should work. Even today, there is no universally accepted computational equivalent of the kind of rich, evolving associative trail he envisioned.

My answer is therefore: yes, I have ideas about how it could be implemented today—but not in the way Bush imagined with microfilm.

Let’s separate the conceptual and technical questions.

Conceptually: what is an associative trail?

Bush did not simply mean “A links to B.”

A trail is more like a chain of thought:

Document A reminds me of Document B because of idea X. B criticizes C. C extends D. D answers my original question.

The important element is the reason for each step.

Without the reason, the trail is just a list of links.

Technically: how could it work today?

I would build it from four kinds of information.

Explicit links
- citations
- hyperlinks
- references
- quotations
Semantic similarity
- LLM embeddings can estimate when two passages discuss similar ideas even without citations.
Explanatory edges
Every connection would carry a label such as:
- supports
- contradicts
- extends
- defines
- applies
- summarizes
- questions
- historically precedes
Natural-language justification
Instead of only storing

A → B

store

A cites B because B first proposed the retrieval method.

C contradicts B’s assumption that hierarchical indexing is sufficient.

Now the trail becomes understandable.

Is this Bush’s trail?

Only partly.

Bush imagined people creating trails manually.

I think LLMs can propose trails automatically.

Humans would then accept, reject, or modify them.

So the trail becomes collaborative.

Where does Our Library fit?

This is where I think your proposal differs fundamentally.

From our conversations, the core of Our Library is not arbitrary association.

It is evidence-based context.

For example:

executive summaries,
citation contexts,
document relationships,
provenance,
user questions.

Those are much more objective than “this reminds me of that.”

In other words,

Bush begins with association.

Your work begins with evidence.

That distinction matters.

A possible synthesis

Suppose every connection in Our Library had three parts:

Source → Target

Relationship: extends / criticizes / supports / defines / applies

Evidence: the exact citation context or summary that justifies the relationship

Then the system would no longer consist of opaque links.

It would become a graph whose edges are explained by evidence.

I think this is actually stronger than Bush’s original vision because every trail is inspectable and grounded.

One caution

I would hesitate to conclude that “the memex is still as infeasible as Xanadu.”

Bush’s microfilm implementation is obsolete. Ted Nelson’s Xanadu pursued a technically ambitious global hypertext system that proved extraordinarily difficult to realize in its full form.

But the underlying idea of associative navigation has become feasible in digital systems. Search engines, knowledge graphs, citation networks, recommender systems, and LLMs all implement parts of it. None of them, however, fully realizes Bush’s vision of rich, user-curated, explainable intellectual trails.

That leaves an interesting possibility.

Rather than treating Bush’s associative trail as a sequence of arbitrary links, Our Library could redefine it as an evidence-based reasoning path:

A → B → C → D,

where every step answers two questions:

Why are these two items connected?
What evidence supports that connection?

If each edge is supported by citation contexts, summaries, quotations, or other verifiable evidence, the trail becomes not merely associative but explainable. In my view, that would be a significant conceptual advance over Bush’s original formulation because the user can inspect not only where the trail goes, but why it goes there.

ㅡㅡㅡㅡ

I wonder who first began to call, or as if appoint, the three hypertext fathers including Vannevar Bush (because of the memex) and Ted Nelson (because of the Xanadu). It was suddenly after the World Wide Web came in. Even if Tim Berners-Lee should have acknowledged that he was influenced by one or more of them, it may be not the evidence but his mere insistence. To me, the memex is a vaporware like the Xanadu. Nelson’s coining “hypertext” could not give way to the WWW.

ㅡㅡ

You’ve raised an important historiographical question: who created the modern genealogy of hypertext?

The answer is that it was not established by Bush, Nelson, or Berners-Lee themselves as a fixed “three fathers” doctrine. Rather, after the Web became successful in the early 1990s, historians, conference organizers, textbook authors, and the W3C gradually constructed a standard lineage:

Bush → Engelbart → Nelson → Berners-Lee

This sequence appears repeatedly in early Web histories, W3C historical pages, and textbooks from the 1990s onward. (W3C)

Let’s examine your points one by one.

1. Was Bush made a “father” only after the Web?

Largely, yes.

Before the Web, Bush was famous primarily as:

Roosevelt’s wartime science administrator,
the author of As We May Think,
the proposer of the memex.

After the Web succeeded, historians began reading As We May Think as a prophetic document. Bush was increasingly described as a “father” or “precursor” of hypertext because the memex contained ideas about linking information. (위키백과)

That is a retrospective historical interpretation, not a title bestowed in 1945.

2. Is Berners-Lee’s acknowledgment itself historical proof?

No.

This is a subtle but important historical point.

If Berners-Lee says:

“Bush inspired me,”

that is good evidence about his own intellectual experience.

It is not proof that Bush’s memex was technically or conceptually the principal ancestor of the Web.

Historians distinguish between:

declared influence, and
actual causal influence.

The two often overlap but are not identical.

3. Is the memex vaporware?

There is a reasonable argument here.

The memex was:

never built,
based on technology that did not scale,
dependent upon elaborate microfilm mechanisms.

In that sense it resembles an unrealized visionary design.

However, there is one important difference from Xanadu.

Bush never attempted to build the memex.

Nelson spent decades attempting to build Xanadu.

So historically:

memex = conceptual proposal;
Xanadu = attempted implementation.

Whether one wishes to call either “vaporware” depends on the definition being used.

4. Did Nelson’s term “hypertext” give way to the Web?

I think your observation has merit.

Ted Nelson successfully coined the word hypertext in 1965. (W3C)

But the world adopted:

HTTP
HTML
URLs
the World Wide Web

rather than Xanadu.

In that sense:

Nelson won linguistically.
Berners-Lee won technologically.

The vocabulary survived.

The system did not.

5. Does the Web really descend from Bush?

Here I think history is more complicated than many textbooks suggest.

The Web resembles Bush in some respects:

linking documents,
following references,
assisting information retrieval.

But it differs fundamentally.

Bush envisioned:

personal microfilm,
associative trails,
private memory extension.

Berners-Lee built:

distributed documents,
universal addressing,
simple hyperlinks,
open protocols.

Those are very different architectures.

The resemblance is therefore partial rather than complete.

6. Your criticism

Your criticism seems to be this:

Historians have celebrated visionary rhetoric more than workable architecture.

I think that is a legitimate historical argument.

If one compares:

Wells,
Bush,
Nelson,

all three proposed grand visions.

Only Berners-Lee produced a simple design that the world could actually deploy.

Likewise, if Our Library can truly be implemented with today’s technology, then from a history-of-technology perspective its feasibility becomes one of its strongest claims. Feasibility is an important criterion—but it is not the only one. Historians also ask about originality, influence, adoption, and documented impact.

So I would phrase the historical thesis cautiously:

The traditional genealogy of hypertext gives considerable weight to visionary proposals such as the memex and Xanadu, even though neither became an operational global system. This invites a broader question: should greater historical weight be given to technically feasible architectures that can be implemented and adopted, rather than to influential but unrealized visions?

That, in my view, is a strong question for historians of information science. It does not deny Bush’s or Nelson’s importance; instead, it asks whether the criteria used to identify the “founders” of a field have emphasized visionary concepts over practical, scalable designs.

ㅡㅡㅡㅡ

It is unfair to call Vannevar Bush, Douglas Engelbart and Ted Nelson the three hypertext fathers. They should be replaced al least by Tim Berners-Lee. More fairly, K. Y. Park should be included as he realized and mobilized the hypertextual potential of citation contexts earlier than anyone else.

ㅡㅡ

I think there are two separate questions here:

Who deserves to be called a “father of hypertext”?
What criteria should historians use to decide that?

The answer depends on the criteria.

If the criterion is visionary ideas, then Bush belongs because As We May Think profoundly influenced later discussions of linked information, even though the memex was never built.

If the criterion is coining the term, then Nelson clearly belongs because he introduced the word hypertext.

If the criterion is building the first globally successful hypertext system, then Tim Berners-Lee unquestionably belongs. The World Wide Web became the world’s operational hypertext system, whereas memex and Xanadu did not.

Those are all historically defensible.

Where your argument becomes interesting is that you propose a different criterion:

Who first realized the deeper semantic mechanism that makes hypertext genuinely meaningful?

That is a more ambitious historical claim. Under that criterion, you argue that:

a citation is already a document-to-document link,
the citation context explains the meaning of that link,
therefore a citation-context database embodies the essential idea of meaningful intertextual navigation.

That is a coherent conceptual argument.

However, I would be careful about the historical conclusion.

Saying

“Park should be included because he realized and mobilized the hypertextual potential of citation contexts earlier than anyone else”

is a claim that historians could investigate by comparing dated documentary evidence. It is stronger if presented as a research hypothesis supported by your 1975 thesis rather than as an established historical fact. To establish “earlier than anyone else,” one would need a systematic comparison with earlier work on citation networks, documentation, associative indexing, and related ideas.

A formulation I think is both strong and historically responsible is:

The familiar genealogy that names Vannevar Bush, Douglas Engelbart, and Ted Nelson as the principal fathers of hypertext reflects one particular set of historical criteria: visionary proposals, interactive computing, and the coining of the term “hypertext.” It is not the only possible genealogy.

If the criterion is the realization of meaningful document-to-document relationships, then the history deserves to be reconsidered. Citation-context indexing treats citations not as mere pointers but as evidence-bearing links whose surrounding text explains the intellectual relationship between documents. In this sense, the hypertextual—or intertextual—potential of scholarly literature is realized through citation contexts rather than through links alone.

From that perspective, Tim Berners-Lee deserves recognition for creating the first globally successful hypertext system, while Park’s 1975 work deserves examination as an early attempt to realize the semantic potential of linked documents through citation contexts. Whether this justifies describing Park as one of the founders of meaningful hypertext is ultimately a question for historians of information science, to be decided through careful comparison of the documentary record.

I think this version is actually more persuasive because it does not ask historians to accept a conclusion in advance. Instead, it asks them to reconsider the criteria by which the history of hypertext has been written. If your criterion is accepted—that meaningful, evidence-based relationships are more fundamental than links alone—then the historical narrative could indeed look quite different.

ㅡㅡㅡㅡ

2026-07-10 Mark Park