By @pagemanSourceOpenVersions

Inspired by this infographic https://x.com/foundinrome/status/2122518660631900371 Teach-me-this-infographic explainer https://chatgpt.com/share/6afc85df-c9f8-83ec-9dc3-298e5b9d892f

Fork
Clone
git clone https://openversions.com/git/pageman/hidden-architecture-of-mathematics.git
Cite
Version link
https://openversions.com/pageman/hidden-architecture-of-mathematics/commit/24dd75f0c6e0f75f817d91f0635a01d1232e41fe
Version identifier (commit hash)
24dd75f0c6e0f75f817d91f0635a01d1232e41fe

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

Claims

Findings

Headline finding

At the level of transitive file imports, the reported Mathlib4 snapshot (commit 3cb72cf, 8,422 files across 27 areas) contains a large mutually reachable core (strongly connected component) spanning 25 of 27 areas (approximately 92.6%), showing that Mathlib's area-level import network is an interconnected formalization ecosystem rather than a clean one-way hierarchy.

Condensed and Information Theory are the two areas outside the 25-area strongly connected component, which under the importer → imported convention means they have no incoming area-level edges from other areas, though this does not mean they have no dependencies of their own or are mathematically isolated.

The highlighted low percentages for Category Theory / Linear Algebra (3%) and Analysis / Category Theory (1%) are file-level import prevalence figures in the indicated direction, not measures of conceptual distance, mathematical unrelatedness, or intellectual influence.

The Set Theory / Dynamics 91% callout in the source infographic uses an arrow shorthand ('SetTheory → Dynamics: 91%') that is potentially ambiguous relative to the chart's row-column convention, and the accompanying note that Dynamics.FixedPoints.Defs 'sits under 50% of mathlib' lacks a defined calculation.

The 33 highlighted upward relationships (orange marks, representing dependencies of at least 20% where the column is 15 or more levels deeper) are a diagnostic signal of dependencies running against the chart's expected ordering, not automatic proof of architectural defects, and the exact level-ordering algorithm cannot be reconstructed from the infographic alone.

A cycle in the area-level import graph does not establish that Lean proofs are logically circular or that every file imports every other file.

Methods

Represent the file system as a directed graph G=(V,E)G=(V,E) where each file is a vertex and an edge f→gf \to g means ff imports gg, with reachability f■gf \blacksquare g denoting a nonempty directed import path from ff to gg, and compute the area-level dependency matrix M[X,Y]=∣{f∈F(X):∃g∈F(Y),f■g}∣/∣F(X)∣M[X,Y] = |\{ f \in F(X) : \exists g \in F(Y), f \blacksquare g \}| / |F(X)|.

Source excerpts

“Let FX be the set of files classified in area X, and let D(Y) be the set of files classified in area Y. Let f ■ g mean that a nonempty directed import path exists from file f to file g. Then the displayed proportion can be represented as: M[X,Y] = |{ f ∈ F(X) : ∃ g ∈ F(Y), f ■ g }| / |F(X)|” · Show in paper

“Represent the file system as a directed graph G=(V,E): each file is a vertex; an edge f→g means f imports g. Reachability f■g means there is a directed path from f to g.” · Show in paper

Interpret the user-supplied Mathlib4 dependency infographic (commit 3cb72cf, 8,422 files, 27 areas) by transcribing its annotations, callout percentages, and legend counts without independently recomputing the dataset or verifying the repository snapshot.

Source excerpts

“This brief interprets the graphic and its annotations; it does not independently recompute the dataset or verify the repository snapshot.” · Show in paper

“The three callout percentages and the 33 upward-relationship count are transcribed from the source image.” · Show in paper

Research files

Commit by @pageman

paper.pdf

Open the manuscript PDF