Demonstration only. Corollary is a non-commercial research demo — no warranty, no support, not for commercial use or clinical decisions. Read the disclaimer

Corollary Labs

Newsletter

The week in science AI, in one email.

New models, datasets and tools for biology, chemistry and materials science come out every day, across a dozen preprint servers and lab blogs. We read them all. On Friday you get the ones that mattered.

The email edition opens shortly. Until then, every release is on the radar and in its RSS feed.

One email, on Friday

The week's releases in the order they matter. Five minutes to read, nothing in between issues, no launch announcements of our own dressed up as news.

What is new, and whether you can use it

The architecture, the data, the benchmark and the number. Then the part release notes bury: is the code out, are the weights out, and under which licence.

Every figure has a source sentence

Each item links to a page listing the claims we made next to the sentence in the paper that supports it. A program checks that the quotes and numbers exist before anything is published.

What an issue looks like

This week, so far.

These are the releases the next issue is being assembled from. Each one opens its full write-up and the evidence behind it.

  1. New tool · Johns Hopkins University

    ALIGNN 2.0: A Dependency-Free PyTorch Rebuild for Materials AI

    A pure-PyTorch rebuild of ALIGNN unifies property, force-field, spectral, and generative prediction in one graph, from Johns Hopkins.

  2. New tool · Anthropic

    Uplifting Biomolecular Modeling: Anthropic's Inference Kits

    Anthropic's 36 drop-in inference kits speed up open protein and genomic ML tools like AlphaFold cofolding, RFdiffusion, and ESM.

  3. New model · Alibaba Group

    LucaCell: A Sequence-Centric Foundation Model for Single Cells

    LucaCell replaces fixed gene IDs with mRNA sequence embeddings, enabling cross-species, cross-modality single-cell analysis without manual mapping.

  4. New dataset · Arc Institute / Tahoe

    Tahoe-100M: a 100 million cell single-cell drug perturbation atlas

    Arc Institute and Tahoe's Tahoe-100M maps 100 million single cells across 1,200 drug perturbations and 50 cancer cell lines.

What happens to your address

It is used to send you this newsletter and for nothing else. It is not shared, sold or added to any other list. We keep the address, the date you confirmed and the wording you agreed to, and no more. Every issue has a one-click unsubscribe that takes effect immediately, and your record is deleted on request to privacy@corollary-labs.com. The privacy policy has the detail.