The work I’m proudest of. tangent/ is an open-source suite that makes scientific computing reproducible and install-free: seven JavaScript modules, built end to end and checked against the reference tools in R and Python. It now has its own home.

suite.tangent.to — the full suite, docs, and architecture.

Why the browser

Almost every device with a browser can already run JavaScript. Put the analysis there and it runs everywhere, with nothing to install and nothing to configure. That is the whole point of the suite: real scientific computing where the audience already is, and results anyone can rerun.

The seven modules

A computational core sits under a higher-level analysis layer:

• Core (MIT): tangent/lina for linear algebra, /proba for probability, /ode for differential equations, /opt for optimization.
• Analysis (GPL-3): tangent/ds, modeled on scikit-learn, for statistics, ordination, and supervised and unsupervised learning; tangent/mc, modeled on PyMC, for Bayesian computation; tangent/sem (experimental), modeled on lavaan, for structural equation modeling.

The results from ds and mc are checked against the reference implementations in R and Python. Full documentation and the architecture map live at suite.tangent.to.

The notebook

The suite runs anywhere JavaScript does, but it has a home of its own: tangent/note, a local-first notebook interface. Notebooks run in the browser, work offline, and your files stay on your machine, with none of the vendor lock-in of hosted notebooks. It started as a tool for algorithmic music composition and grew into a general home for accessible scientific computing. The suite also runs on Observable and on the Node.js and Deno runtimes.

A short course on Observable walks through the workflow end to end: data science with tangent.