How scoring
works

The index carries scores. Best carries a personal recommendation.

Digna legi is Latin for “worth reading”—the judgment this site exists to make.

A program watches 206 hand-picked sources. Articles are scored against a written reading profile using titles, summaries, and retrieved text where available. The evidence is sometimes brief: a score is not a claim that the full article, its research, or its visuals have been reviewed.

Evidence before judgment

The system follows a simple epistemic rule: uncertainty should remain visible. Evidence-reviewed evaluations use retrieved article text when available, with excerpts clearly identified. A new or revised score is accepted only when both the judgment and its explanation can be traced to supporting passages and survive a separate model audit. When the evidence runs out, the process stops; it defers judgment rather than disguising ignorance as a low number.

Most published scores have now passed this process. The few that have not remain explicitly marked Historical score—not because the article failed, but because the evidence or validation did. English selection notes face their own evidence check; if a revision fails, the previous safe wording remains. These checks reduce unsupported claims; they do not make model judgment infallible or rewrite Ratipcan’s permanent Best selections and historical reading issues.

The profile favours lasting value over novelty. Relevant evidence, useful recommendations, depth, cross-domain connections, and clear mechanisms count. Repetition, clichés, superficial hype, and digressions that add little work against a piece. Length alone earns no credit.

Reading briefs

Selected pieces have a short AI-written main idea and an optional deeper reading. The deeper layer preserves useful mechanisms, distinctions and examples from the source. Its length follows the material: a compact essay does not need a padded framework. When useful, a clearly labelled editorial question helps you think with the idea.

A separate model checks new layered briefs for source support, specific detail, distinct contributions, substance, proportionate treatment and clarity. Supporting passages are required for deeper claims and examples. Introductions and incomplete previews cannot stand in for missing treatment. This reduces unsupported claims but does not guarantee accuracy.

Technical reading labels describe prerequisites, separately from the numeric score. Summaries appear first for featured and newly selected writing; pieces without a checked summary still open with their selection note and original link.

What the bands mean

• 80–100High value: lasting insight, strong synthesis, or an unusually good explanation.
• 70–79Worth the time, with more limited depth or contribution.
• < 70Below the usual publication threshold. Personal Best selections are an exception.

What it is not

This reflects one reader’s preferences and the evidence available when scoring. It is not an objective quality rating. Scores can change as the profile or available evidence improves.

Best is the editor’s permanent selection. Its infinity mark is a personal distinction; the underlying score stays unchanged.

Where the articles come from

Most are found by scanning selected sources. Some come from the editor’s reading archive and other published reading lists; those carry a line of credit under the title.