Serendip Commons SocietyTowards Relational Being ← All projects

The Science Infrastructures

The sciences as one connected body of knowledge, to be learned, used and extended.

An open, integrative infrastructure for learning, using and researching all domains of science, in theory and in practice. Foundation libraries carry mathematics, computation, physics and inference. Each science is a domain: a library of its own objects and models, pure and tested, that runs on its own with only the domains beneath it. Applications and studies, organised by purpose, compose the domains they need. Each part can grow, change or be replaced without disturbing the others.

Domain I · Theoretical & Ontological Domain VI · Educational Domain II · Epistemic MIT License Active SCS/PA/2026/05

View on GitHub ↗ The full gallery ↗ Project Approval (PDF)

The sky, a history, and a random geometry: three of the infrastructure's ten applications.

Structures shared by the sciences

The sciences are taught and practised apart, yet the same structures recur across them. Symmetry: groups act on molecules and crystals, on the sphere of the sky, on quantum states and on sequences, and their representations give spectra and harmonics. Conservation and change: the same equations transport heat in the ocean, chemicals in a cell and populations across a landscape. Networks: food webs, neural circuits, reaction networks and economies are graphs with dynamics on them. Sequences and grammars: genomes, languages and programs are strings with structure. Inference: every field estimates hidden states from noisy observations.

An infrastructure that holds the sciences together makes these connections explicit. It shares one implementation of each method, tests it once, and reuses it everywhere. The sciences are taken in the broad sense of the German Wissenschaft, the social sciences and the humanities among them: a language is a grammar, a chronology is a reference frame, a market is a network.

Three tiers

Works	Applications and studies, organised by purpose, each composing the domains it needs.
Shared edges	Data (a registry, acquisition, products, requests on demand), visualization, observation and sensors, and an API through which other programs reach the infrastructure.
Domains	One library per science, pure and tested: fields, and bridges between them.
Foundations	Mathematics, computation, physics and inference, written once and used by every domain.

The domains are ordered by what each may depend on. The order is acyclic, and each domain passes its checks alone with what lies beneath it:

height 4   climate
height 3   tides · navigation · air–sea · planetary science
height 2   astronomy · geophysics · geography · oceanography · meteorology
height 1   geodesy · evolution
height 0   metrology · neuroscience · history · genomics · economics · robotics
           ──────────── foundations: physis ─► calx, morphe · tyche

The foundations

calx

Computation: a retargetable stack and numerical methods, from linear algebra, integrators and special functions to task graphs, vector lanes and real-time scheduling.

morphe

Mathematical structures and their equivalences: exact arithmetic, groups and their characters, complexes and their homology, words, grammars and their parsers.

physis

Geometry and physical law: units, mechanics from a discrete action, rotations and frames, fields as sections, lattice gauge theory, spin foams and group field theories.

tyche

Chance and inference: distributions, the description of samples, regression, time series and spatial statistics.

Three activities

Every artefact says which it serves: learning (derivations, worked examples, expositions), use (tools whose numbers can be relied on) or research (open questions and new results).

Verifiable

Numbers are held to authoritative references, among them SOFA and ERFA, JPL Horizons, the IERS Conventions and GeographicLib, each with its source and date; theory is checked by computation where it can be.

What it draws

Ten applications compose the domains. These pictures are from the repository's gallery, where each caption says what is shown, where its data come from and how it was made. Select a picture to see it at full size.

The Earth Atlas

The atmosphere, the ocean, the land, its people and the solid Earth on one globe or on equal-area maps: forecasts from several models and two of its own, the ocean beneath any place, the plates and their motion, earthquakes, the crust, and climate from 1901 to 2099.

The tectonic plates coloured by speed, with their boundaries and moving particles — The tectonic plates, coloured by speed.

Wind at 10 m over the North Atlantic on the globe, with moving particles — Wind at 10 m over the North Atlantic.

A section across the Japan Trench with earthquakes tracing the sinking Pacific plate — Across the Japan Trench, the sinking plate traced by earthquakes.

Köppen–Geiger climate classes over Asia, 1991–2020.

The Celestial Atlas

The Sun, the Moon, the planets, the stars and the galaxies at any instant from 1849 to 2150, and the sky from any place on the Earth.

Galaxies of the 2MASS Redshift Survey linked into groups — Galaxies of the 2MASS Redshift Survey, linked into groups.

The sky over San Pedro de Atacama, chosen on a small globe — The sky over San Pedro de Atacama.

The Brain Atlas

The human brain in the round, every part with its anatomy; sections, atlases, the head and its electrodes, the mouse brain, and EEG, MEG and fMRI carried onto the cortex.

The deep structures of the brain seen through a glass cortex, with the hippocampus chosen and its anatomy on a card — The deep structures through a glass cortex, the hippocampus chosen.

The alpha rhythm of a resting person with closed eyes, estimated on the cortex from 64 EEG channels — The alpha rhythm of closed eyes, estimated on the cortex from 64 EEG channels.

An fMRI map of listening to stories against arithmetic on the cortex — An fMRI map of listening to stories.

The Spiking Simulator

Networks of spiking neurons from twenty-two models, classical and quantum, run by the neuroscience domain's C++ engine with the same spikes on any number of threads.

A sixty-millisecond close-up of the raster in the fast oscillation, the spikes in narrow vertical volleys — Brunel's network in its fast oscillation.

Firing rate against current for three models: a continuous rise for two, a jump for the third — The classes of excitability: a continuous rise, and a jump.

Photons of driven and pulsed quantum emitters as a raster — Photons of quantum emitters, read as spikes.

The Complex Atlas

Cell complexes with their exact invariants and the gauge fields on them, and the generation of spacetime by field theories: from an action, through its Feynman graphs and spin foams, to the condensed phase and the continuum.

A spin foam of Boulatov's group field theory with one vertex opened onto its spin network and quantum tetrahedron — A spin foam, one vertex opened onto its quantum tetrahedron.

A U(1) gauge field on a lattice torus, its links coloured by angle — A U(1) gauge field against the exact two-dimensional value.

The volume of a group-field-theory condensate bouncing in relational time — A condensate whose universe bounces.

The History Complex

Histories built into 2-complexes: occasions are vertices, the lives of beings edges, and each tie between two beings a 2-cell. Their loops, shells and signed ties are read exactly, the ties as the spin foam of a ℤ₂ gauge theory fitted to the record. Its canons are seeds drafted by a language model and marked unchecked until checked against their sources.

Europe 1785–1828 as a foam of lives and ties rising through time, cut at 15 November 1814 — Europe 1785–1828 as a foam of lives and ties.

Vienna 1890–1938 as a foam, its lenses bright where they are likeliest in a spin foam — Vienna 1890–1938, bright where a spin foam is likeliest.

The loom of Florence and Rome 1401–1564, focused on the tie of Pope Julius II and Michelangelo — The loom of Florence and Rome, 1401–1564.

The Genome Workbench

Genomes mapped, aligned, indexed, read, assembled and placed on their trees by a C++ engine on the genomics and evolution domains, with a toolbox of instruments for one's own sequences and files.

Escherichia coli K-12 on the circle, with the strip opened around its origin of replication — *E. coli* on the circle, opened at its origin of replication.

Two reads aligned by the wavefront algorithm, its fronts drawn over the edit graph — An alignment's wavefronts on the edit graph.

The tree of cytochrome b of ten primates by maximum likelihood with bootstrap support — Ten primates by maximum likelihood.

The Economy Atlas

Value carried between places and through time: the world's trade split into exchange, drive and circulation; currencies held and exchanged as a spacetime whose curvature is the departure from interest parity; production as a circular flow; and the loop deficits of transfer networks, which vanish exactly when a system of prices exists.

The world's trade in goods in 2024, countries coloured by their attraction, the largest net flows as arcs — The world's trade in 2024.

Twenty currencies against the euro, 2015 to 2026: the curvature of every monthly lens — Twenty currencies as a spacetime, the curvature of each month.

A network of five positions and six legs, its two chords' holonomies and the potential of its tree — Five positions, six legs, and their holonomies.

Cosmopoiesis

A framework for constructing and simulating physical worlds. Its first world is the Solar System, from a crater's floor to the orbit of Pluto, on a console drawn as a precision instrument: its fields drawn in the world, the ground saying where each of its heights comes from, the Earth beneath any place to its centre, and instruments whose every reading says whether it was observed, modelled or assimilated.

The Earth's magnetic field lines coloured by the field's strength — The Earth's magnetic field lines.

A section through Japan to 800 km at true scale, the earthquakes tracing the sinking plate — Beneath Japan to 800 km, at true scale.

The Moon's ground near Hadley Rille coloured by where its heights come from — The Moon's ground, coloured by where its heights come from.

The screenshots are the infrastructure's, under its MIT licence; the data shown in them keep their own licences, listed in the gallery's credits.

Where it stands

As of 30 October 2026, the repository's README records:

4foundation libraries with C++

16domains with code, of fifty-five mapped

10working applications

7,472checks in thirty-five suites

The domains of the Earth and the sky are held to SOFA, ERFA, JPL Horizons, NAIF, GeographicLib, PROJ and Vallado's verification set; history, genomics, evolution, economics, planetary science and robotics are C++ from the start; each domain writes no general method of its own, but uses the foundations'. The next steps are set out in the repository's roadmap, and every decision, with its reasons, in its decision log.

Run it yourself

You need Git, Python 3.12 or later and a current browser; nothing else for the three atlases. The first run downloads about 450 MB of public data.

git clone --recurse-submodules https://github.com/huangwanhong-maker/The-Science-Infrastructures.git
cd The-Science-Infrastructures
python3 -m venv .venv
.venv/bin/python -m pip install -r data/requirements.txt
.venv/bin/python data/acquisition/recipe.py applications/earth-atlas/recipe.json
.venv/bin/python applications/earth-atlas/serve.py --open

On Windows, write .venv\Scripts\python. The applications with a C++ engine (the Spiking Simulator, the Complex Atlas, the History Complex, the Genome Workbench, the Economy Atlas and Cosmopoiesis) also need g++, in WSL on Windows. The Quick Start gives every application's lines, and the public instances of the three atlases.

Free to use, free to continue

Open source. Every repository is released under the MIT License. Anyone may use, study, change, sell and redistribute it, fork it and carry it on, with or without the project and with or without the Society.
The copyright stays with its makers. Each contributor keeps the copyright in what they contributed. The Society is the project's steward and holds no copyright or other intellectual property in it; it cannot relicense it.
Contributing. Contributions are made under the same MIT License, to everyone, with a sign-off (git commit -s) certifying the Developer Certificate of Origin. There is no contributor agreement. Each repository's CONTRIBUTORS.txt holds the terms and the register of contributors.
AI assistance, stated. Much of the code was written with AI coding assistants, and the commits say so. They are recorded as tools; the person who submits the work is its contributor and answers for it.

Governance

Approved by the Founder as SCS/PA/2026/05 (30 October 2026), as sole Director during the Initial Period under §14.5 and §14.6 of the Bylaws, and announced by the notice SCS/NTC/2026/02. Classified under Domain I (Theoretical and Ontological), with Domains VI (Educational) and II (Epistemic) as secondary. Under the Society's Intellectual Property Policy (SCS/POL/2026/01) the Society retains nothing; the technical direction stays with the project's maintainers. Ethics review under §3.5 is not required at the present scope, which involves no human subjects and no personal data; any work with personal data, and the planned work on drug discovery, will be examined before it begins.

Read the Project Approval (PDF) · Read the Notice (PDF) · The project's own record of the decision

⏱ Both instruments are blockchain-timestamped (OpenTimestamps), so their integrity and date can be verified independently of this site: approval proof (.ots) · hash (.sha256); notice proof (.ots) · hash (.sha256). Verify with ots verify project_approval.pdf.ots.