The ideas are Iñigo González de Arrieta's, shared with Josu M. Igartua, who develops them here. The known case is powder diffraction: a Rietveld refinement is compatible with many sets of atomic coordinates, and mode crystallography — the same structure in symmetry-adapted coordinates — is what tells them apart. The proposal is to make that move general: write the expert's knowledge into scripts, treat massive data sets with them, and learn over the result what no single analysis can show. This site holds the brief, a discourse that sorts the ideas, three cases side by side, a programme of proposals, the sources with their status, and a dated log.
The brief says the experiment collapses the configuration space. The discourse splits that one word into three things that need three different remedies: what the forward map cannot see at all (powder averaging, the phase problem), what it sees too weakly to pin down against noise (the long valleys of sloppy fits, the curved credible region of Daun et al.'s Fig. 9), and what it cannot represent because the model is wrong (Rayleigh against Mie; a free-electron metal that liquid germanium is not). Symmetry modes, priors, other experiments and model-error terms each answer one of the three.
Findings, decisions with their reasons, and the work done. Newest first.
sources/library/pdf/{daun-…, kuzmenko2005, transtrum2015, perezmato2010}.pdf re-read 2026-10-14 · L009–L011ecosystem/tools/audit_page.py · ecosystem/site/sources/citation-audit-2026-10-14/source.md · project.css · L010The founding note is on the Brief page verbatim and organised into numbered theses, none left out. Every later page says which theses it develops.
Iñigo's, Josu's, or this development's: the Discourse marks the boundary at each step, so that the discussion knows what it is discussing.
Every reference on the Sources page says whether it was read here, confirmed by a search, or is cited from memory.
tools/build.py rebuilds the navigation, the theses, the log and the sources from data/*.yml on each render. Decisions go in the log the day they are taken.