How your papers are chosen & scored

Stage 1 · Gathering

Every morning we pull the newest papers and preprints from OpenAlex (which indexes most scholarly journals), arXiv, SocArXiv and PsyArXiv — searching the research areas learned from your seed papers. Each candidate is then compared to your seeds by semantic closeness (meaning, not keywords), and the closest form your daily shortlist.

Stage 2 · Selection — the judge

An AI reader (a large language model) reads every shortlisted title and abstract against your Selection Criteria — the plain-English description you wrote, plus your flavors, exclusions, and example papers — and scores fit for you:

9–10 squarely inside one of your flavors — the flavor is the paper's central question
7–8 clearly within a flavor, sharing the stage with other aims
4–6 competent on one component (your topic area alone) without the combination
1–3 tangential — shares vocabulary, not the research space
0 off-target or on your not-interested list

Every score comes with a one-sentence rationale — that's the quoted line on each card. If the rationale is wrong, that's your cue to edit your criteria.

What are "flavors"?

Named sub-interests — each one an intersection of things you care about (e.g. "narrative_persuasion: narrative as a persuasion mechanism, in any domain..."). A paper squarely inside one flavor is a bullseye; it doesn't need to touch all of them. The tags on each card's chip show which flavors the judge thinks a paper engages. Edit them in Settings → Selection Criteria.

Your feedback is part of the system

Thumbs-up and thumbs-down on any card become boundary examples shown to the judge from the next run onward — "papers the researcher liked / rejected". A few votes on borderline cases sharpen your radar noticeably. Clicking through to a paper is also recorded so we can evaluate how useful your briefings are.