Colophon

How this site works, in brief. The full colophon, with the taxonomy and the rules for curated links, is in Italian.

Languages

The site holds two bodies of work under one premise. The research programme on validating language model outputs is in English: the programme page, the Lab, the papers, the measuring bench. The essays, the curated links, the concept index and the map of the site are in Italian. The two languages do different work; the Italian archive is not a translation waiting to happen.

The AI-involvement notation

Every published text ends with a code declaring how far a language model was involved in producing it. The code measures how much of the published surface passed through a model, not the intellectual weight of the contribution. Responsibility for every published text is mine.

First axis: the prose. Always present, one value, cumulative in the permissive sense.

00

None. No use of language models beyond the spellchecker.

DL

Dialogue. Preliminary discussion, brainstorming, research and fact-checking conducted in conversation with a model. The text is mine from the first line.

ED

Editing. Text written entirely by me, then revised with assistance.

WR

Writing. First draft generated from my instructions and my materials, then rewritten and edited by me.

Second axis: the apparatus. Optional, combinable.

FM

Formalisation. Assisted construction of technical apparatus — formulas, models, code, tables — appearing in the published text.

TR

Translation. Assisted adaptation from or into English of a text already mine.

The notation was introduced on 23 August 2026; codes on earlier pieces are declared reconstructions. The specification is a versioned page of its own, released under CC BY 4.0 so that anyone can adopt it: the specification (in Italian), and the machine-readable vocabulary in notazione.json.

The map

The map on the home page is generated at every publication from the concept index. A concept can carry a geography: one or more countries or regions, and the way it concerns them. Geographic fields are assigned and validated by the author when a piece is published; there is no automatic extraction.

For each concept with a geography, the linked pieces are counted and multiplied by 1 if the concept is a place or a country, 0.5 if it has a direct geography but is not a place, 0.25 if the link to that country is only theoretical. Values are summed country by country. The European Union has a weight of its own that also colours its member states; a member with references of its own adds the two. Regions give full weight to each country they include. The five colour classes are quantiles of the legend values. The figure next to each country is therefore a weight, not a count of mentions. The Italian side uses the same data under a different projection (Lambert azimuthal equal-area, centred on Bergamo).

Data

The concepts assigned to each piece form a dataset published as JSON-LD at concetti.json, with terms documented at /ns/, under CC BY 4.0 and deposited on Zenodo: https://doi.org/10.5281/zenodo.22843731.

Licences and technology

The essays are released under CC BY-NC-ND 4.0: they can be shared with attribution, for non-commercial purposes and without changes. The downloadable teaching materials in the Learning Log are under CC BY-NC-SA 4.0; the concept dataset under CC BY 4.0. Static site built with Eleventy and published on Netlify; typeface Source Serif 4, served from this domain. No profiling cookies, no paywall, no recommendation algorithms. Statistics: Google Analytics 4, aggregated, only with consent.