puckworks

Evidence-first models for what happens inside an espresso puck. Papers become typed components, components face validation gates, and every conclusion keeps its limits.

GitHub-only distribution — not on PyPI. Source and releases: github.com/trbrewer/puckworks.

Explore

Explore why the cup hides the clock

A static, generated interactive: many hidden inventory–rate combinations can produce nearly the same final espresso cup, so a good endpoint fit can hide weak parameter identification. Every number is generated from the committed Paper A result bundle.

Open “The Cup Hides the Clock” →

See how we falsified our own espresso headline

A static, generated interactive: an attractive interior “fine-grind” extraction-yield maximum weakened once we rebuilt the target on one coherent TDS basis and plotted the raw replicate points. The corrected cell means are ordered with the middle dial below the coarse dial — no interior maximum — while a model can still draw such a feature without identifying it. Every number is generated from the PV-04 producer and bound to the CC-BY Schmieder source files.

Open “How We Falsified Our Own Espresso Headline” →

See why espresso model components do not simply add up

A static, generated interactive: an extraction-only model tracks the rising flow, but adding a swelling branch through one shared porosity state over-closes it and makes the prediction worse than a flat line. It rejects one tested composition — not swelling in general. Every number is generated from named producer functions.

Open “More Physics Made It Worse” →

Explore the component registry (Google Colab)

Run the CPU quickstart: the registered components, their provenance, and the validation gates (installs the released wheel after verifying its SHA-256).

Open the quickstart in Colab →

Run a guided espresso pull (Google Colab)

Enter a bounded recipe and run one coherent, model-backed pull stage by stage, with evidence-labelled traces and honest domain warnings. It does not model puck wetting, physical first drip, a dynamic pressure profile, temperature response, or flavor.

Open the Guided Espresso Pull in Colab →

Evidence and reproduction

This site is a small landing page; the repository README remains the project homepage. No analytics, no tracking, no external scripts.