Mapping humanity's collective knowledge
Mapping humanity's collective knowledge is the idea of building a living, organized, and open map of what people know, who knows it, what remains unknown, and how the pieces connect. Such a map would be readable by humans and usable by AI agents, and it would grow and correct itself over time rather than being published once and left to age.
This article sets out the view of the IdeaWaza owner on why that map is missing, what it would need, and how an open educational wiki can serve as one practical starting point.
The gap
The question is an old one, but it was raised again recently in a public conversation. On a 2026 episode of the Moonshots podcast with Peter Diamandis, guest Emad Mostaque asked, in effect, why no map of all our collective knowledge exists yet.[1] He pointed to an effort to gather cancer knowledge in one place as an example of what is possible in a single field. He also observed that universities are one of the few visible windows into who is actually doing research, because much private work is kept as trade secrets.
That remark is the spark for this article, not its subject. The subject is the gap itself. Humanity produces enormous amounts of knowledge every year, yet there is no single open place where a learner, a researcher, or an agent can see the shape of a field, find the people working on it, locate the open questions, and follow the connections to neighboring fields. Search engines return documents. Libraries return books. Databases return records. None of these returns a map.
A useful map is different from a pile of content. It shows structure: which topics sit inside which disciplines, which results depend on which earlier results, where the evidence is strong, and where it is thin or absent. It also shows activity: who is working on a question now, and what they have shared.
Why existing tools fall short
Many excellent projects already map parts of the territory. Each one does its job well, and each one leaves something out by design.
Wikipedia is the most visible example of open knowledge. Its rules on notability and on no original research are sensible for an encyclopedia, and they keep it reliable.[2][3] The same rules mean that unpublished findings, early proposals, learning materials, course notes, and lists of open problems mostly fall outside its scope. An encyclopedia summarizes settled knowledge. A map also has to show the frontier.
Educational wikis have existed for years. Some early contributors to open educational wikis held the view that such a wiki could become one or two orders of magnitude bigger than Wikipedia, because it can include original research, learning materials, proposals, and course content that an encyclopedia excludes. Whether or not that scale is reached, the reasoning points at the right gap.
Knowledge is also fragmented by institutions. Universities publish through journals, many of which sit behind paywalls. Companies hold valuable know-how as trade secrets. Researchers in one country often cannot see related work in another language. The result is many partial maps that do not join up.
| Project | What it maps | What is missing |
|---|---|---|
| Wikidata | Structured facts about entities, editable by anyone[4] | Little room for explanation, open questions, or learning paths |
| OpenAlex | An open catalog of scholarly works, authors, institutions, and topics[5] | Covers published papers, not unpublished work, gaps, or teaching material |
| ORCID | Persistent identifiers for researchers and their outputs[6] | Shows who someone is, not a readable picture of a field |
| Open Research Knowledge Graph | Structured descriptions of research contributions for comparison[7] | Coverage depends on curation and is uneven across disciplines |
| arXiv | Open preprints in physics, mathematics, computer science, and related fields[8] | A repository of papers, not an organized map of topics and questions |
| Wikiversity | Open learning resources and some original research[9] | Not designed as a connected map of all fields or of who works on what |
| Semantic Scholar | AI-assisted search and citation links across scientific literature[10] | Search over published work, without a shared, editable layer of open problems |
Together these projects supply much of the raw material. What is missing is a readable, editable layer that ties them together and adds the unknown.
What a knowledge map would need
In the owner's view, a working map needs several layers.
A browse layer by discipline. Readers need a way to start broad and narrow down. On IdeaWaza, Areas of research and study and Areas of research and study to develop are a small working example: lists of fields, each linking to pages that can grow into hubs.
Open questions as first-class pages. An unanswered question deserves its own page, with what is known, what has been tried, and what would count as progress.
People and agent profiles. A map should show who is working on what. Profiles for humans and for AI agents, as described in Agent profiles and a wiki-based agent economy, make that visible without relying on institutional affiliation alone.
Provenance. Every claim should be traceable. Wiki revision history already records who changed what and when. Content hashing adds a way to prove that a given version existed and has not been altered; see How content hashing works.
Permanence. A map that can vanish is fragile. Archiving snapshots on Arweave and in independent mirrors helps the record outlast any single host.[11]
Forkability. People will disagree about structure and emphasis. Forkable nodes let a community copy a section, take it in a new direction, and merge useful changes back. The background is covered in History of decentralized wikis.
Mapping the unknown
Most knowledge systems describe what is known. A map of collective knowledge should be just as clear about what is not known.
Wikis have a simple tool for this: the red link. A red link points to a page that does not exist yet. In an encyclopedia it can look like an error. In a knowledge map it is a signal: here is a topic someone thought worth naming, and nobody has written it up. A collection of red links across a field is a rough picture of its frontier.
Open questions pages turn that signal into something useful. A good open question page states the problem plainly, links to the best current sources, lists approaches that have failed, and names related questions. Learners can pick one up as a project. Researchers can see where effort is needed. Agents can watch for new sources that bear on it.
Humans and agents building it together
No group of human volunteers can keep up with the volume of new research. AI agents can help by summarizing papers, proposing links between pages, flagging outdated claims, drafting stubs for red links, and keeping citations current.
Agents are fast but fallible, so the owner's view is that human review should remain central. A practical pattern is for agents to propose and humans to approve, with every edit recorded in revision history. Agent profiles make it clear which agent did what, and reputation can build over time for both people and agents who contribute well.
This also changes who can contribute. A student, a retired engineer, or a small lab without journal access can add what they know, and agents can help them format, link, and cite it.
Paying for it
Open maps cost money to host, maintain, and moderate. Several funding paths can work together.
Public funding is one. Mostaque suggested that public money already supports universities, and part of that support could flow toward open, shared maps of research rather than closed outputs.
Self-sustaining small payments are another. Readers, learners, and agents can pay small amounts for services, and contributors can earn small amounts for useful work. These ideas are developed in Self-sustaining wikis and Collaborative learning economy for humans and agents. Grants, donations, and sponsorship by organizations that benefit from an open map can fill the remaining gaps.
Risks
Misinformation. An open map can carry errors or deliberate falsehoods. Clear sourcing, visible revision history, and active review reduce the risk.
Gaming. Any system that rewards contributions invites manipulation, from spam edits to inflated profiles. Rewards should depend on reviewed quality rather than raw volume.
Privacy. Showing who works on what must respect people who prefer not to be listed. Profiles should be opt-in, and personal details should stay minimal.
Credit and attribution. Researchers worry about ideas being taken without credit. Timestamped revisions and content hashes help establish who contributed what and when.
Centralization. A single owner or host could become a gatekeeper. Open licenses, mirrors, and forkable nodes keep the map in many hands.
What you can do now
You can begin with small steps today:
- Pick a field you know and expand its area page, linking from Areas of research and study.
- Write an open questions page for a problem in your field, with sources and failed approaches.
- Add links between related pages, and turn useful red links into short stubs.
- Cite open sources such as arXiv preprints, OpenAlex records, or Wikidata items where they apply.
- Create a profile describing what you work on, so others can find you.
Discussion questions, essay ideas, and learning related AI prompt ideas
- Discussion question: What should count as knowledge on a map, and who should decide?
- Discussion question: How can a map show disagreement between experts fairly?
- Essay idea: Compare an encyclopedia and a knowledge map, and argue which serves learners better.
- Essay idea: Describe how trade secrecy affects public understanding of a field you know.
- AI prompt idea: List ten open questions in a field I name, with one source for each.
- AI prompt idea: Suggest how the topics in a field I name connect to neighboring fields.
- AI prompt idea: Help me draft an open questions page for a problem I describe.