Max Fürst

New preprint🚨 Imagine (re)designing a protein via inverse folding. AF2 predicts the designed sequence to a structure with pLDDT 94 & you get 1.8 Å RMSD to the input. Perfect design? What if I told u that the structure has 4 solvent-exposed Trp and 3 Pro where a Gly should be? Why to be wary🧵👇

Finally looks like rollout is really starting! They indicate a timeline of 7 days, then Mythos and Opus 5.5 (but not Fable) with relaxed bio safeguards

Life scientists get in line, our turn to help build the new generation of agents that'll replace us. Raise your hand if you can resist the temptation of those sweet sweet tokens and quick splashy papers this might bring

did a group presentation today with manim style transitions. Sharing the template for anyone to build upon Brought to life with manim-web + reveal.js + claude + codex view template: slides.ramith.fyi view repo: github.com/ramithuh/pro...

The fear mongering is starting to get ridiculous "Easier said than done" - uhh duh! www.theguardian.com/technology/2...

In response to the government’s renaming of AI to SI, I propose all journals rename Supplemental Information to Additional Information.

1/5 Cryo-EM gives high-resolution RNA structures, but only when high Mg²⁺ freezes out their motion. At physiological conditions, RNAs are dynamic and 3D reconstruction fails. Can we still learn their folding equilibria? New preprint: yes, straight from the raw particle images👇

images that spring to mind unbidden, part 1007: "more hydrogen atoms"

„Without funding of basic research – which is the advantage of doing research in Germany – our discoveries would not have been possible”, Nobel Prize laureate 2026 Peter Hagemann points out in German news @heutejournal.bsky.social (starting at 19:00 minutes) – in German www.zdf.de/play/magazin...

3D printers have now become zero-skill infrastructure when connected to an Ai. OpenSCAD + PrusaLink + Claude is a miracle in my lab. Highly recommend it. I am free.

Visualising the distribution of substrings in the genomes of different organisms. Going up to depth eight by recursively splitting into quadrants, we get these fractal-like pictures.

Thoughts on Anthropic’s worries about people accessing Claude for possible bioweapons design:

Join to find out more about our ADD-tagging platform plus a preview on our soon-to-be preprinted follow up presented by my PhD student @wahyuwidodo.bsky.social

Hate to conclude this prediction was correct

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...

Anthropic has used Claude to optimise structure prediction and design code to improve speed and memory efficiency. 2-3X speed improvements on AF3, 5X on OF3, 5X on BindCraft low memory modes to model large assemblies. All released as open source www.anthropic.com/research/cla...

'Our findings illustrate a curious paradox: despite universities being key producers of evidence and seeking the use of their research to inform public policymaking, they are less adept at using evidence to inform their own decision-making.' blogs.lse.ac.uk/impactofsoci...

Thought: If scientific journals start requiring authors to declare the use of AI in any part of their work (as they absolutely should!), then they should also require the same declaration from *reviewers* 🧵1/3

How do researchers design an RNA in 3D when no accurate 3D prediction model exists? Eterna's collaboration with the @rdaslab.bsky.social at Stanford was able to design novel pseudoknots using a blend of secondary structure prediction and GenAI design methods. 🧪 #RNAsky

Extraordinary blog post by @moalquraishi.bsky.social moalquraishi.wordpress.com/2026/08/10/a... Starting with the question on whether workaphiles will be dying breed post AGI, it later turns into serious scifi - "The 'easy' solution is to make our brains bigger"

A multivalent RNA affinity tag enables selective purification of tagged RNAs and bound proteins doi.org/10.1093/nar/...

Shocking.

A couple paid more than $800,000 for a gene-editing therapy for their daughter. She died, and it wasn’t made public. An exclusive from us and @science.org.

This "scalable" template for oral examination requires 20 minutes per student per exam, or 67 total hours for a class of 200 students. It would appear the authors and I have a different understanding of what "scalable" means. www.insidehighered.com/opinion/care...

Ironically, as @mart1oeil.eu showed, there are telltale signs that the piece was edited with ChatGPT: many links include "?utm_source=chatgpt.com" If even Science editorial team cannot afford good human editing, that's worrying

1/10 🚨Preprint! One RING to rule them all (phages)👑💍⚔️ Bacteria build giant DNA-scaffolded RNase rings (larger than the ribosome!) to shut down viral infection🧬🧵👇 biorxiv.org/content/10.6...

Couldn't care less about football but love meself a random chart

🚨 We're hiring! My new lab at EMBL Hamburg is looking for a Research Technician. I'm looking for someone excited by phages, metabolism, enzymes, structural biology and bacterial immunity—if you know the right person, please share this post!🧵⬇️ Apply here: embl.wd103.myworkdayjobs.com/en-US/EMBL/j...

The @qedscience.bsky.social "impact" score generated a lot of discussion on ranking preprints, including ideas on multi-dimensional rankings that are user specific. Besides describing the ideas, we can now prototype them (with Claude in this case). Here is Salient salient-sgeu6fs5ua-uc.a.run.app

Not Claude advertising their new model to me using a left-handed helix

Please note the image is from AI and is not scientifically accurate. We intended to add visual appeal to the post, and apologize for posting this fanciful but unfortunately misleading depiction. In future posts, we will ensure scientific images are based on actual data.

Roche throws the axe 🪓 at illumina with Axelios open.substack.com/pub/albertvi...

Can't believe it Sent a cryo dewar with crystals via #FedEx to @esrf.fr and they lost it. Since 2 weeks gone without a trace and of course they're doing nothing to try find it. Equipment worth €8000, dozens of priceless xtals from our projects gone 4ever 😭 Compensation? €600 from FedEx (by weight) 😡

Beautiful videos nicely explaining microtubules! Congratulations to Renaud Chabrier and Casten Janke 👍 www.youtube.com/watch?v=dVbv...

Since the 1960s, the genetic code has been used to predict protein sequences from DNA and mRNA sequences. Our @nature.com article demonstrates that these predictions miss thousands of protein sequences present in human tissues. www.nature.com/articles/s41... 1/

Here is an example of what I expect we will start seeing more of, a single author bioinformatics paper that looks like it was made with AI assistance. The figures look very much straight out of Claude Code. It does have a "Use of AI tools" disclaimer section journals.plos.org/ploscompbiol...

wtaf biology, as if you hadn't had enough enzyme mechanisms for adding oxygen already

Unconventional monooxygenation by the O2-dependent tRNA wobble uridine hydroxylase TrhO www.nature.com/articles/s41... 🦠 cryo-electron microscopy structure of Bacillus subtilis TrhO–tRNAAla complex

I wrote about AI in academia. "PhD-level thinking", LLM bias, grunt work, alignment, AGI, data center water use, AI politics -- something for everyone.

Grep this: Microsoft grafts (most) Linux commands onto Windows

well if you liked this you might also be interested in the graph I made for the same course on the abundance of elements. Green are the CHNOPS element, which constitute 98% of living matter on earth. Fun to speculate about the ones way off the axis