Release notes

A running log of what we've changed and why it matters for your weekly digest. We ship continuously; this page is updated whenever a change reaches all subscribers.

Release 2026.04.19

Latest

Sharper relevance, cleaner digests, and a faster, more reliable Monday delivery.

  • Improved

    Sharper relevance for your weekly digest

    We upgraded the part of our system that understands what a paper is actually about, going beyond simple keyword overlap to a deeper, meaning-based comparison with your research interests. In practice, you should see fewer borderline matches and more papers that genuinely belong on your reading list — particularly in interdisciplinary areas where wording varies between fields.

  • Improved

    Non-English papers are now filtered out

    Some indexed feeds occasionally include papers whose title and abstract are not in English. We now detect this — both for non-Latin scripts (e.g. Chinese, Arabic, Cyrillic) and for Latin-script languages (e.g. Spanish, French, German, Portuguese, Italian) — and exclude them before ranking. Your digest stays in English, with no manual cleanup required.

  • New

    Smarter pre-flight checks in the Keyword Advisor

    When you create or edit a project, the Keyword Advisor now performs a richer health check on your keyword set before your first digest. It flags terms that are too broad, too narrow, internally contradictory, or written in a language other than English, and it explains each flag in plain language. The goal is simple: catch problems while you can still fix them, not after Monday morning.

  • Improved

    Project digests now surface up to 10 papers (was 4)

    For paid subscribers, each research project in your account now receives up to 10 ranked papers per week instead of 4. Several of you told us that 4 felt like a teaser; 10 gives a meaningful weekly read while still being a finished, triaged list rather than a stream.

  • Improved

    Cleaner deduplication across journals and preprint servers

    The same paper is sometimes listed by more than one source (for example, an arXiv preprint that later appears in a journal). We normalised how paper identifiers are compared so duplicates and near-duplicates are now collapsed into a single entry. Expect fewer "didn't I just see this?" moments inside a digest.

  • Reliability

    Faster and more resilient Monday delivery

    Behind the scenes we made the weekly fetching pipeline substantially faster and added automatic retries when our data sources are under load. The customer-visible effect: your digest is much less likely to be late on Monday morning, even on weeks with heavy publishing volume or database pressure.

  • Fixed

    Several backend stability fixes

    We resolved a handful of edge cases that could cause a small number of subscribers to miss a paper, see an empty section, or receive a delayed digest. None of these were widespread, but if you noticed any of these patterns recently, they should be gone now.

Spotted something off in your digest, or want to suggest an improvement?

Get in touch →