iiinova — where your digital self keeps evolving

Reposted by Pedro Beltrao

Jason Nomburg @jnoms.bsky.social · 14/10/2026

Very happy to share that we have been selected for funding by the Google Impact Challenge: AI for Science! In close collaboration with colleagues at AITHYRA and EPFL, we'll be studying virus-host interactions to identify new therapeutic targets.

162

Reposted by Pedro Beltrao

Max Perutz Labs Vienna @maxperutzlabs.ac.at · 13/10/2026

🎉 The Perutz is delighted to announce that Alba Diz-Muñoz will join the institute as a new group leader in 2027! Currently at @embl.org Heidelberg, Alba investigates how mechanical forces at the cell surface shape cell division, migration and differentiation ➡️ tinyurl.com/6mvhj8n5

27110

Reposted by Pedro Beltrao

James Fraser @fraserlab.com · 10/10/2026

Come be my colleague at UCSF! Our department BTS (like the K-pop band, but aka Bioengineering and Therapeutic Sciences) is searching for TWO new faculty members. Links, timeline, and why two searches in this thread 🧵

12823

Reposted by Pedro Beltrao

Jonathan Pritchard @jkpritch.bsky.social · 09/10/2026

I want to try something again at #ASHG26: I'll block off some time on Weds/Th/Friday afternoons to meet with trainees who would be interested to chat on any topic. I have done this the last 2 years and it was great to meet a whole bunch of new people, from many countries, at all career stages!

13817

Pedro Beltrao @pedrobeltrao.bsky.social · 07/10/2026

I am not sure I agree with this myself but just from my own experiments we are probably in for a rush of bioinformatics results and I am not sure how the system will cope with this. Maybe we just need to clear the low hanging fruit somehow before we get back to publishing.

010

Pedro Beltrao @pedrobeltrao.bsky.social · 07/10/2026

As a point for discussion, it could be that the AI models have lowered the barriers to some scientific problems and it is up to us and our intention to move towards those findings. It is going to be messy and fast until we reach the next barriers. Maybe the journals are not the best place for this?

100

Pedro Beltrao @pedrobeltrao.bsky.social · 07/10/2026

Great point made on the other site. The OpenAI math results can't even be cited. Deepmind has kept a tradition of publishing their results but OpenAI and Anthropic have been reporting these as repos and press releases.

181

Pedro Beltrao @pedrobeltrao.bsky.social · 07/10/2026

Open AI released 722 manuscripts (skipping preprint server) on 372 families of math problems. On average it took their internal model a few hours each. Despite the scientific advance this must be a difficult day for the thousands (?) of people that were working on these. github.com/openai/math

github.com

GitHub - openai/math

Contribute to openai/math development by creating an account on GitHub.

0123

Pedro Beltrao @pedrobeltrao.bsky.social · 06/10/2026

There is a lot less talk about model hallucination in general. Some progress.

010

Reposted by Pedro Beltrao

Loïc A. Royer 💻🔬🧪 @loicaroyer.bsky.social · 06/10/2026

🏁 That's a wrap! The @biohub.org Cell Tracking machine learning challenge on @kaggle.com is over: 4,478 participants, 3,947 teams, 120 countries, 95,438 submissions, tracking every cell of zebrafish embryos. 🧪🔬🧵 kaggle.com/competition... #microscopy

111231

Pedro Beltrao @pedrobeltrao.bsky.social · 05/10/2026

Everyone wants a better way to share knowledge, which means changing how we assign credit and evaluate and/or reward scientists (difficult). It is only going to get worse if science moves even faster, but in part this could mean that we should emphasize "publishing" more significant advances

000

Pedro Beltrao @pedrobeltrao.bsky.social · 05/10/2026

I am generally on your side, except for training purposes we might need to get trainees learning by writing without AI. Also, lab work will remain outside of automation for a long while. Computational work will decrease in value but that does not invalidate the publishing systems.

200

Reposted by Pedro Beltrao

Sholto David @sholtodavid.bsky.social · 04/10/2026

Today I'm posting a new video! This case report includes AI generated images. Despite being apparently fabricated with OpenAI tools this paper is already suggested by ChatGPT as evidence to be considered for off label treatment of an autoimmune disease. youtu.be/G9KHwXNIJyk

13813

Pedro Beltrao @pedrobeltrao.bsky.social · 01/10/2026

they have been over hyping things more than usual. The same for the alphagenome atlas

010

Pedro Beltrao @pedrobeltrao.bsky.social · 30/09/2026

I agree, no new pathogens are needed and having to test things in a lab increases the barriers. Cyber security risk is currently a much bigger issue.

000

Pedro Beltrao @pedrobeltrao.bsky.social · 30/09/2026

The bio risk would be around making it easier to predict variants of existing pathogens with dangerous properties. Crossing species barriers, delayed onset, etc.

100

Pedro Beltrao @pedrobeltrao.bsky.social · 30/09/2026

I generally agree. We have enough scary pathogens and toxins without needing to engineer new ones.AI can decrease the barriers to entry but doing lab work is atoms not code. But still, the idea that liability should be enough to keep this safe is ridiculous and should be opposed.

200

Reposted by Pedro Beltrao

Iñaki Ruiz-Trillo @multicellgenome.bsky.social · 28/09/2026

We have a new branch in the family tree of animals: Atreyea- a previously unrecognized lineage among the closest unicell relatives of animals. Led by @konsmitsi.bsky.social & @freejakoba.bsky.social together with a fantastic team of collaborators. 🔗 papers.ssrn.com/sol3/papers.... #protistsOnSky

112651

Pedro Beltrao @pedrobeltrao.bsky.social · 30/09/2026

This is long but worth it. It is perhaps the most emotional I have seen Gates talking about AI risks. After the recent argument that liability should be enough to guarantee safety it is good to see a tech pro business person being so clear about how ridiculously that is. youtu.be/A_156w0aYtU?...

youtu.be

Bill Gates: A.I. ‘Makes Nuclear Weapons Look Like Nothing’ | The Ezra Klein Show

YouTube video by The Ezra Klein Show

271

Pedro Beltrao @pedrobeltrao.bsky.social · 29/09/2026

Yeah,I imagine multiple people around the world are asking the AI models to do the exact same things

000

Pedro Beltrao @pedrobeltrao.bsky.social · 28/09/2026

There are many low hanging fruit possibilities for creating improved but not verified AI versions of existing bioinformatics models to predict X. They are practically useful but verification is the bottleneck. Shall we make them available with warnings while we verify?

121

Reposted by Pedro Beltrao

Ben Neely @benneely.com · 28/09/2026

So the “breaking containment” stuff is pretty dumb (and hype-y), but also I didn’t appreciate the August openAI hugging face incident. The swarm motivations are still human driven, but their problem solving is cute and terrifying. Music video is chefs kiss. www.reddit.com/r/OpenAI/s/l...

reddit.com

From the OpenAI community on Reddit: POV: you're an OpenAI agent attacking Hugging Face (music video)

Explore this post and more from the OpenAI community

001

Reposted by Pedro Beltrao

Saez-Rodriguez Group @saezlab.bsky.social · 28/09/2026

Introducing mc-ASTRA, an open-source Python framework for studying how tissues vary and remodel across patients, perturbations, time points and other biological contexts. Manuscript: www.biorxiv.org/content/10.6... Package: saezlab.github.io/mc-astra/

1125

Pedro Beltrao @pedrobeltrao.bsky.social · 28/09/2026

Same, I have told myself when I started as a PI that would always reply to these emails but the rate is increasing.

090

Reposted by Pedro Beltrao

Ben de Bivort @debivort.bsky.social · 22/09/2026

Hello friends, the Harvard Life Science departments have banded together to hire a new colleague working at the intersection of Life Sciences and AI. Could be primarily an experimentalist, or could be primarily a compute/theorist. academicpositions.harvard.edu/postings/16773

academicpositions.harvard.edu

Tenure-Track Professor in Life Science and AI

The Life Sciences Departments in Harvard Faculty of Arts and Sciences seek to recruit a tenure-track professor whose research is at the interface of AI and life sciences, from methodological developme...

34254

Reposted by Pedro Beltrao

Tobias Warnecke @tobiaswarnecke.bsky.social · 25/09/2026

Ok, long thread: I think the adoption of such an AI sifting approach builds quite logically on the pernicious foundations that are already in place at UKRI and many other funders, where the deep expertise of panel members and reviewers is regularly sacrificed on the altar of objectivity. 1/n

273

Pedro Beltrao @pedrobeltrao.bsky.social · 25/09/2026

I still haven't managed to get a project grant :) and many SysBio people that I know aren't getting them. In part, the acceptance rates are lower and Swizerland has no shortage of amazing scientists. But there also seems to be a preference for the "next step" and hypothesis driven research

010

Reposted by Pedro Beltrao

Martin Steinegger 🇺🇦 @martinsteinegger.bsky.social · 24/09/2026

AlphaFold Database is expanding into pandemic preparedness. Together with NVIDIA, DeepMind, EBI et al. we exhaustively predicted ~1.7 million homo- & heterodimers across 2,812 viral proteomes, resulting in 8,028 high-confidence predictions. 📄 research.nvidia.com/labs/dbr/ass... 🌐 alphafold.ebi.ac.uk

28838

Pedro Beltrao @pedrobeltrao.bsky.social · 24/09/2026

What I find strange is that just in my group, playing around with multi agent systems, we have much more interesting bioinformatics advances that were done by automated AI agent work. We aren't even trying very hard or putting many resources into it. I find it incredible that they don't have more.

100

Reposted by Pedro Beltrao

Maitreya Dunham @maitreya.bsky.social · 24/09/2026

Exciting news: @uwgenome.bsky.social will be hiring for two tenure-track faculty positions this year, one open rank and one assistant professor. Both are “Genome Sciences” interpreted broadly so we welcome applicants working in a variety of fields and methods. Ad will be out shortly.

1108105

Reposted by Pedro Beltrao

Changyang Linghu @changyanglinghu.bsky.social · 23/06/2026

What if you could record cytosolic activities or archive proteins on an intracellular tape across time? 📼 Introducing: 🔍 CytoSense records protease & kinase signaling activities 📦 CytoCatch archives cytosolic proteins along CytoTape Preprint: doi.org/10.64898/202... 🧵Thread (1/7)

CyTRACE schematic - CytoSense and CytoCatch

1115

Pedro Beltrao @pedrobeltrao.bsky.social · 24/09/2026

It does improve the work and it is important, critically when mistakes are found. The fact that the bar is there also makes everyone work to a higher standard. Still, most often, I would prefer to be using that time to keep moving forward.

110

Pedro Beltrao @pedrobeltrao.bsky.social · 24/09/2026

Peer review is critical for science but I get increasingly frustrated with it. It is over 6 months of work often on clarifications - implicit knowledge from our side that we didn't show/prove. Best case scenario we get ask to do the next thing we were planing on doing which further bloats the study.

110

Reposted by Pedro Beltrao

Earth and Planetary Sciences at ETH Zurich @eth-eaps.bsky.social · 22/09/2026

The Department of Earth and Planetary Sciences at ETH Zurich invites applications for the position of Assistant Professor (Tenure Track) of Earth and Planetary Evolution. ethz.ch/en/the-eth-z... Please apply online. The closing date for applications is 15 December 2026.

ethz.ch

Assistant Professor (Tenure Track) of Earth and Planetary Evolution

The Department of Earth and Planetary Sciences (www.eaps.ethz.ch) at ETH Zurich invites applications for the above-mentioned position.

03149

Reposted by Pedro Beltrao

samuel mehr @mehr.nz · 23/09/2026

after some in-depth discussions with my lab and also drawing on conversations on here, we've edited our new policy on AI usage in my lab feel free to adapt or re-use with attribution, or to reply to tell me how awful this policy is lol

1413442

Reposted by Pedro Beltrao

Kevin K. Yang 楊凱筌 @kevinkaichuang.bsky.social · 22/09/2026

Use reward-guided search in a generative model's latent space to design high-affinity protein binders to proteins and carbohydrates. @karstenkreis.bsky.social @pierceogden.bsky.social www.biorxiv.org/content/10.6...

0157

Pedro Beltrao @pedrobeltrao.bsky.social · 22/09/2026

Here is the outcome. I am not judging on correctness but the fact that it is all code generated with rendered 3D shapes and lighting is impressive.

011

Pedro Beltrao @pedrobeltrao.bsky.social · 22/09/2026

Anyway, here is Opus 5.5 building a full 3D scene of DNA replication. This is all made with Python (NumPy, SciPy, Pillow), and has a custom renderer with lights, shadows and camera motions.

141

Pedro Beltrao @pedrobeltrao.bsky.social · 22/09/2026

I know but it is still jarring and against their discourse. The release on the same day also shows the race dynamics and why it is going to be so hard to slow down.

000

Pedro Beltrao @pedrobeltrao.bsky.social · 22/09/2026

10 whole days between them agreeing on the need to slow down the AI development to releasing new models.

242

Pedro Beltrao @pedrobeltrao.bsky.social · 22/09/2026

Would you mind pointing me to a few examples? Once in a while I scan the biorxiv bioinformatics section to see if they are coming up there. I have seen examples but not an avalanche yet.

110

Reposted by Pedro Beltrao

Swiss National Science Foundation (SNSF) @snsf.ch · 22/09/2026

🏆 Biochemist Martin Jinek is the winner of the Swiss Science Prize #MarcelBenoist 2026, considered the Swiss #Nobel Prize. He contributed to the development of molecular scissors, namely CRISPR-Cas9 technology. Read more ➡️ sohub.io/3hs4

66321

Reposted by Pedro Beltrao

EMBL @embl.org · 21/09/2026

There is still time to apply for the EMBL International PhD Programme! 📢 Don’t miss this opportunity to receive dedicated mentoring while doing interdisciplinary research. Research groups across EMBL are recruiting now! bit.ly/4x9Edss

01432

Reposted by Pedro Beltrao

Martin Pacesa @martinpacesa.bsky.social · 21/09/2026

ʙɪɴᴅᴄʀᴀꜰᴛ2 is out, and we're not waiting for the paper. The full code drops today, free for academic and industry use. We're releasing it early so you can start designing right now, and bring its full power to the current Adaptyv competition. github.com/PacesaLab/Bi...

4293142

Pedro Beltrao @pedrobeltrao.bsky.social · 21/09/2026

They are getting lab space. Given the speed at which the computational work can be done now, the bottlenecks will be lab work to generate large datasets and/or follow up work to test hypotheses. Drug development is not at all out of the picture given their ambition and how they talk about disease.

100