Ariel Ben-Sasson retweeted
🚨 RFdiffusion3 is officially published in Science!
From the Baker lab, this next-generation AI model enables atom-level biomolecular design, going beyond protein structures to engineer complex catalytic sites.
Experimentally validated results include de novo enzymes that degrade organophosphates and cleave disease-related peptides.
📄 doi.org/10.1126/science.aeh2…
🤖 Made with AI
Ariel Ben-Sasson retweeted
Finally a more intuitive way to learn pLDDT/pAE? 😎
sokrypton.github.io/protein_…
(Character idea from @HannesStaerk & Alex Waldherr)
Ariel Ben-Sasson retweeted
We've made a breakthrough in self-evolving AI scientists moving from "search" to "principled discovery": Scientific discovery requires that the search space itself changes, and an AI scientist must perceive this shift without intervention. We built an AI that achieves this for the first time with the ability to discover the scientific vocabulary it reasons in. Evidence, tools, artifacts, verifiers, failures & claims become typed provenance. We show three distinct modalities: 1) retrieval, adding known objects; 2) search, exploring a fixed schema; and critically: 3) discovery, a verified regime transition.
We solve the open-endedness evaluation problem by lifting agentic workflows into a typed copresheaf and proving, via a Kan obstruction, that true discovery is not unbounded generation but a verifiable schema expansion: old evidence is transported by Left Kan extension, and genuine novelty is mathematically quantified by the pointwise residual beyond the transported image - separating discovery from mere search and making novelty objective and measurable rather than a subjective judgment or benchmark delta.
Our AI scientist is built in a way that does not pre-conceive the approach it chooses; instead, we endow the system with formal power to adapt, evolve, and reason from first principles. Case studies include:
1⃣Builder/Breaker model that discovers mode-conditioned compliance in proteins;
2⃣CategoryScienceClaw that finds anisotropic fiber-network stiffness rules.
Great work in collaboration with my graduate student @fwang108_ @MITdeptofBE
F.Y. Wang & M.J. Buehler, Self-Revising Discovery Systems for Science: A Categorical Framework for Agentic Artificial Intelligence, arXiv:2606.01444, 2026
Ariel Ben-Sasson retweeted
We're delighted to kick off our new seminar series, and the inaugural speaker is none other than Neil King @KingLabIPD
"ML-based design of nanoparticle vaccines"
June 3rd 2026
12-1 PM EDT
Ariel Ben-Sasson retweeted
The IPD is looking to hire a full time on site Public Information Specialist!
wd5.myworkdaysite.com/en-US/…
#claudecode has been the most transformative day-to-day tool in science & engineering recently, #protein engineering in my case... Hitting quota became both my hobby and my "please not now" moment, so I've let #claudecode build a little menu bar app for MacBooks to track itself.
1
2
164
One command install:
> brew tap Arielbs/claude-usage-monitor github.com/Arielbs/claude-us… &&brew install --cask claude-usage-monitor
> GitHub: github.com/Arielbs/claude-us…
1
106
Rosetta MCP server
#Rosetta, #RosettaScripts, @PyRosetta , and @RosettaCommons docs are now accessible as modular components in agentic workflows. Translate XML scripts into PyRosetta code, let your agent reason on mover functionality, and more in your MCP enabled @cursor_ai IDE
You can find it here:
🔗 GitHub: lnkd.in/gXib8Mtp
📦 NPM: lnkd.in/gtaiFZGy
2
160
I’m excited to give a Short Talk about protein compression and drugs generation at @KeystoneSymp #MachineLearning Applied to Macro#MolecularStructure and Function, this March! Looking forward to meet old and new friends and colleagues and explore emerging research in Keystone!
(Non-scientific update!)
Broke my high-altitude (3,000 meters) 10K snow-run record today here in Keystone! Only minutes after presenting our work in the “Machine Learning for Protein Design” session. 🚀❄️🏃♂️ #ProteinDesign #MachineLearning #KeystoneSymposia
Ariel Ben-Sasson retweeted
📢Excited to introduce NanoCas -our new mini CRISPR system that can reach tissues previously out of reach!
By shrinking CRISPR to 1/3 its normal size, we can now edit genes in muscle, heart & brain that were difficult to access before. Summary & link to paper:
Ariel Ben-Sasson retweeted
Best read of the day - innovative design, and great science to generate large virus-like protein cages by breaking symmetry 🎉
nature.com/articles/s41586-0…
nature.com/articles/s41586-0…
nature.com/articles/s41467-0…
Quinton Dowling,
@lsmin0152
@kribler
@KingLabIPD
@UWproteindesign
Ariel Ben-Sasson retweeted
1/🧬 Excited to share PLAID, our new approach for co-generating sequence and all-atom protein structures by sampling from the latent space of ESMFold. This requires only sequences during training, which unlocks more data and annotations:
bit.ly/plaid-proteins
🧵
Ariel Ben-Sasson retweeted
Chai-1 has always been available for commercial use via our server. Today, we're also making Chai-1(r) code and weights available under an Apache 2.0 license, which permits broad commercial use.
github.com/chaidiscovery/cha…
Ariel Ben-Sasson retweeted
Genie 2 matches the best existing methods on designability while opening comfortable margins on diversity and novelty, likely by implicitly capturing more of protein structure space. Despite being trained on crops of <256 residues, Genie 2 holds its own up to 500 residues. (2/5)



























