Emily

Founder @ Ligo Bio

[Arda Göreci @ArdaGoreci]

Excited to share more of our technical notes in the spirit of open-research. Blog 2: How Muon Lost its Geometry

As Muon spread from speedruns to LLM-scale, it lost the µP scaling rule that keeps the best LR as width changes. The bug ships in PyTorch and Optax 😱. (1/n)

[Emily @eggywarbs]

@ArdaGoreci Brought to you by our protein cowboy

[Emily @eggywarbs]

@ArdaGoreci Nature experienced the og mode collapse

[Arda Göreci @ArdaGoreci]

Today, we are launching our research blog!

We’ll use it for technical notes from our work building tools for enzyme and biomolecular design.

Our first post is about The Unreasonable Redundancy of Nature's Protein Folds.

TLDR: Please don't fold more sequences (1/n)

[Ed Harris @edharris]

Disappointing to see @Nature fail to acknowledge that all teams who reproduced DeepMind's code (Ligo Biosciences, @chaidiscovery, @Baidu_Inc, ByteDance) found significant errors in the pseudocode which prevented training and compromised reproducibility. https://t.co/RbOuXk7yNL

[nature @Nature]

“We’re very excited to see what people do with this”

AlphaFold3 is open at last https://t.co/cdvACli8r5

[Ed Harris @edharris]

Excited for Ligo Biosciences to be included in both @Nature Magazine and @ScienceMagazine in their recent articles on the release of @GoogleDeepMind 's AlphaFold3 code. https://t.co/xliW6LTjBE

[Sergey Ovchinnikov @sokrypton]

And this is why code should always be published with papers... This group made an attempt to reproduce AlphaFold3 and found a number of potential issues in the published pseudo-code. See exciting 🧵 👇

[Arda Göreci @ArdaGoreci]

🚀Excited to announce: Open-source AlphaFold3 implementation! 🚀

I am thrilled to announce one of the models we have been building for the last 8-weeks at Ligo - an open-source implementation of DeepMind’s frontier model, AlphaFold3! Here’s what we have learned, a thread (1/11):