Cladus the genealogy of research
Phylogeography of Science

Where ideas begin,
and where they spread

An open data pipeline tracing the lineage, migration and divergence of research on the citation graph — seeded at Bell Labs' terahertz work (Auston, 1975). Start anywhere and Cladus keeps handing you the next thing.

Data OpenAlex · arXiv Seed Bell Labs THz · 1975 Method citation trails + reading lists License MIT (planned)
Papers
Researchers
Institutions
Citation links
Topics
Year range

A stream that learns what you read

Every paper you open teaches Cladus. It clusters what you keep returning to into mixes, and any paper can seed a station that keeps going.

Your reading lists

Reset taste

Researchers like you

Closest topic mix to yours — and one thing each works on that you haven't

In your stream

Each card says why it's here
Loading…

Following the trend, or filling a gap?

The same graph that recommends things can also place you on it. Every topic you read gets two coordinates — how crowded it already is, and whether its share of new work is growing — and your reading is the weighted average.

Reading your profile…

Search the corpus

Every paper, researcher and institution collected from the seed outward. Open one to follow its lineage — or start a station from it.

Reading the landscape

Founders and eras, crowded fields and untouched combinations, and the next wave as the model sees it. Every number is computed from the collected citation graph.

A phylogeography of research

This project treats the evolution of science with the toolkit of population genetics and phylogeography. The metaphor is not decoration — it is the method: phylogeny estimation, migration flows, and founder-effect detection, applied directly to the citation graph. What follows is what that method actually returned on this corpus — the branches, the crowded ground, the blanks — followed by how it was built and what it cannot see.

How the line branched

Descent from the 1975 seed, collapsed to the lines that carried enough papers to stand on their own.

Where the work concentrates

Papers placed by what they are about, not by which topic label they were filed under.

The empty spaces

Two different kinds of absence — a blank the field has grown around, and a pair of neighbours that never cite each other.

Enclosed blanks

Cells with no papers, ringed on most sides by cells that have them. The margin of the map is empty for boring reasons; these are not.

Under-linked neighbours

Spaces the geometry places close together, citing each other less than volume alone would predict. Lift below 1.00 is the signal.

Method

How the numbers above were produced.

Concept mapping

Concepts from population genetics translate into structures of the citation graph.
GeneResearch idea / technique
InheritanceCitation & co-authorship
MigrationSpread across institutions & countries
MutationConcept change / topic drift
Founder effectA few seeds dominating a field
AdmixtureField fusion (THz × medical imaging …)

Pipeline

A four-stage open pipeline, standard library only.
01
Collect
BFS from the seed papers across citation neighbors — papers, authors, geocoded institutions, topics and the citation graph into SQLite — collector/
02
Analyze
Lineage trees, founders, migration trajectories, concentration & gaps, growth forecasts — analysis/
03
Recommend
Citation trails (co-citation, coupling, topic overlap, authorship) and topic clustering from the event log — analysis/recsys.py
04
Visualize
Feed, newsroom, 3D globe, analysis dashboard — web/
Data sources
OpenAlex — primary arXiv Nature RSS Crossref

The seed — Bell Labs' terahertz lineage

The lineage roots at Auston's picosecond photoconductive switch (1975). Collection starts from these five works and expands the citation graph in both directions.
YearWorkRole
1975Picosecond photoconducting switch — D. H. AustonRoot of the lineage
1984Picosecond photoconducting Hertzian dipolesOrigin of THz emission
1988Subpicosecond photoconducting dipole antennasThe antenna line
1988Electrooptic generation/detection of femtosecond pulsesThe electro-optic line
1990Generation of femtosecond EM pulsesGeneralized fs pulses
Recommendations are computed from the citation graph, never from a black box — every card in the feed carries the reason it appeared. Your event log stays local, in data/events.db, and “Reset taste” erases it. Coverage, name disambiguation and language biases are documented as known limitations.

Models

Three of them, all standard library, none downloaded — so every number on this page can be recomputed from the same database.

Where a language model would earn its place

A design, not a deployment. Nothing on this page is written or ranked by a language model today.

The geometry above has a specific weakness, and it is worth naming before proposing a cure: it reads a paper as a bag of words, so it cannot tell a paper that measures a thing from one that reviews it. Both say “terahertz spectroscopy” and both land in the same neighbourhood. That is the gap a language model would close, and it suggests four jobs in the order they would pay off.

Claim extraction. Read each abstract into a structured triple — material, method, quantity measured — and cluster on those instead of on vocabulary. A research space would then be defined by what people do, which is what the word was supposed to mean.

Naming. A cluster currently called “conductivity · characterization” is a list of tokens the statistics liked. A model reading the twenty most central papers could give it the name a researcher in the field would use, with those twenty papers cited as the reason.

Bridge triage. The under-linked pairs above are candidates, and most candidates are coincidence — two areas can sit close in word-space and stay apart for an excellent physical reason. An agent with retrieval over both sides could sort the accidents from the real openings, and say which it is.

A question interface. Tool access to this same database — “who moved from photoconductive antennas into topological materials, and when” — answered by running queries, not by recalling training data.

Three constraints would apply to all four. Every generated sentence carries the work ids it was derived from, and a sentence without ids does not render. Generated text is stored in its own table, marked as generated, and never joined into atlas.db without a human pass — a corpus that quietly mixes collected and inferred records is no longer evidence. And the site stays fully functional with the model switched off, which is also why it is off by default: the figures above are reproducible by anyone holding the same database, and an API-backed sentence is not.

Following