iiinova — where your digital self keeps evolving

Reposted by Erik Thiede

pilarcossio.bsky.social @pilarcossio.bsky.social · 10/02/2026

#BPS2026 The T&C subgroup is hosting a fun event on Saturday evening, Feb 21st 🎉 Open bar for 2 hours + food! Tickets are selling fast, so sign up soon: www.biophysics.org/2026meeting/...

biophysics.org

BPS2026 Annual Meeting Subgroup Dinners| Biophysical Society

123

Erik Thiede @erikhthiede.bsky.social · 09/01/2026

(Just looked at the comments and saw that @liweichang.bsky.social beat me to it)

110

Erik Thiede @erikhthiede.bsky.social · 09/01/2026

That's unfair! Sometimes it's a fast-folding protein without solvent :p.

110

Erik Thiede @erikhthiede.bsky.social · 04/01/2026

To be pedantic it just "calls a tool that returns info". That tool *could* be a web search, but it could be something else. For instance, if your LLM is focused on sports, you could have it consult a database of baseball scores directly -- you don't need to pass through a web search.

020

Erik Thiede @erikhthiede.bsky.social · 04/01/2026

RAG queries an external tool (e.g. the raw data from a search), compress its content into a numeric representation, and incorporate that alongside your prompt. Another closely related technique is CAG (cache-augment generation). CAG works on fixed data sources, but is typically much faster.

140

Erik Thiede @erikhthiede.bsky.social · 04/01/2026

Good question. There are two options here. The first (naive) option is you keep retraining the model. Obviously this is slow, expensive. The other option is do basically what you described. The keyword to look up here is "RAG": Retrieval Augmented Generation.

130

Reposted by Erik Thiede

Sjors Scheres @sjorsscheres.bsky.social · 10/12/2025

Postdoc opportunity with Michel Goedert and myself to work on #cryoEM/#cryoET structures of #amyloids. Feel free to email me with any questions you might have. mrc.tal.net/vx/mobile-0/...

mrc.tal.net

Postdoctoral Scientist - Neurobiology - Dr Michel Goedert - LMB 2752 - Medical Research Council

Location: Cambridge. Vacancy: Postdoctoral Scientist - Neurobiology - Dr Michel Goedert - LMB 2752. Closing Date: 04/01/2026, 23:55

12123

Erik Thiede @erikhthiede.bsky.social · 05/12/2025

Really interesting work. Is there a github somewhere we can play around with?

000

Erik Thiede @erikhthiede.bsky.social · 05/12/2025

Using some matrix perturbation theory, we can build heuristics for which MSM states are sensitive to error: we find that states at transition states or in metastable regions on pathways can lead to large errors.

010

Erik Thiede @erikhthiede.bsky.social · 05/12/2025

When I started working on MSMs I was told never to trust rate estimates from MSMs -- the errors were just too big. Now, almost a decade later (jeez I'm old) I think we have some ideas around where this error comes from.

100

Erik Thiede @erikhthiede.bsky.social · 02/12/2025

I'm giving a talk at One World Cryo EM (cryoem.world) tomorrow at 11:00am EST. Topic is 'Deconstructing Algorithms for “Discrete” Heterogeneity in Cryo-EM'. Sign up for their mailing list to tune in!

cryoem.world

042

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

It leaves me feeling conflicted when we mention him as senior scientists, so I thought I'd say something. But if others feel differently about him and interpret his character differently, that's entirely their prerogative :).

000

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

The reason I commented is that when we tell the next generation to turn to Feynman for inspiration we are also teaching them how to "be" in science. As such it is also worth discussing the bad with the good.

200

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

I absolutely do not want to take away from anything that gave anyone inspiration to pursue science. On a personal level I have considerable respect for your work, and I am glad this inspiration occurred.

100

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

And you do see it in this interview as well: like when he talks about social science.

100

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

I think it's that while Feynman's attitudes towards women were quite problematic, I think it's part of a broader pattern that underlies some of his stories: other people are there for Feynman to use.

100

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

Right, there's a good question on why I'm not just posting by myself about Feynman's attitudes towards women (I'd certainly not be the first to do so!) And certainly one thing I would never want to take away from the inspiration we find as young scientists.

100

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

He then is curious if it works with a "nice, ordinary, Southern girl" so he tries it "with the sister of a grad student." And he says it works, and he sleeps with her. But presumably he realizes this is all somewhat distateful, so he decides to stop doing it because "he didn't enjoy it."

000

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

He learns to pick up girls by, "adopting the attitude that those bar girls are all bitches, that they aren't worth anything, and all they're in there for is to get you to buy them a drink, and they're not going to give you a goddamn thing; I'm not going to be a gentleman to such worthless bitches"

200

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

He's really an amazing storyteller and they're all charmingly told. You really start rooting for this rebellious underdog who comes in with his simple questions that upset the scientific apple cart. But halfway through you get to the part where he starts talking about going to bars.

100

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

My own experience really starts with having read "Surely your Joking" Mr. Feynman, a set of interviews that serve his memoirs, which I was asked to read as part of an undergrad chemistry class.

100

Erik Thiede @erikhthiede.bsky.social · 23/11/2025

I wish I could feel the same way. He's an amazing storyteller, but the more you read between the lines, you see how often he is using his stories to whitewash his own sexual harassment or condescend to the people around him.

100

Erik Thiede @erikhthiede.bsky.social · 18/11/2025

www.thecrimson.com/article/2025... ...and what's more, absolutely none of this happens in a vacuum. There's much I love about academia, but one thing I absolutely hate is the way it permits professors to abuse the power they have over their mentees. We have systemic issues we need to face.

thecrimson.com

As Summers Sought Clandestine Relationship With Woman He Called a Mentee, Epstein Was His ‘Wing Man’ | News | The Harvard Crimson

When former Harvard President Lawrence H. Summers was pursuing a romantic relationship with a woman he described as a mentee, he turned to a longtime associate for guidance: convicted sex offender Jef...

040

Erik Thiede @erikhthiede.bsky.social · 29/10/2025

(Jokes aside: package looks great! Thanks for putting this together.)

000

Erik Thiede @erikhthiede.bsky.social · 29/10/2025

Request: I think with a little bit of effort, we could further improve dimensional efficiency and make py1Dmol. I've attached a quick proof of concept in the image. Happy to submit a PR if you want to follow up.

170

Erik Thiede @erikhthiede.bsky.social · 25/09/2025

I almost think it's inherent in the field. I think we are a "wastebasket taxon" for all the things that didn't fit in Biology and Physics 😅.

110

Erik Thiede @erikhthiede.bsky.social · 22/09/2025

We are hiring for a postdoc to work on the intersection of molecular simulation, machine learning, and cryo-EM! Please find the job posting here. thiedelab.github.io/docs/postdoc...

thiedelab.github.io

021

Erik Thiede @erikhthiede.bsky.social · 26/08/2025

Back from our first group conference trip to ACS Fall. Was great to not only see new science, but also excellent watch the very excellent presentations from Josh (Rhodes), Diego, and Jeffrey, as well as the great reception Josh (Almonte)'s poster received!

010

Erik Thiede @erikhthiede.bsky.social · 17/07/2025

Excellent read for anyone trying to understand what is up with all the AI hype nowadays.

010

Erik Thiede @erikhthiede.bsky.social · 08/06/2025

And here I thought it was if you try *non*-symplectic integrators you don't go back.... to the same energy value.

120

Erik Thiede @erikhthiede.bsky.social · 29/05/2025

Posting a job ad on behalf of Mat Sikora, who is doing some fascinating work on cryo-EM for glycans. He's looking both for PhDs and postdocs. science.phd.uj.edu.pl/join/biomedi...

science.phd.uj.edu.pl

Szkoła Doktorska Nauk Ścisłych i Przyrodniczych - Uniwersytet Jagielloński

000

Erik Thiede @erikhthiede.bsky.social · 26/05/2025

Happy to see our first collaboration with the Yang group come through!

031

Reposted by Erik Thiede

James Fraser @fraserlab.com · 20/05/2025

We are looking to hire (yes, even in this economy!) a jr. specialist to train in protein prep/structural biology related to our AVOID-ome work as part of openadmet.org. A great position for someone who is looking to be a tech for a few years before grad or med school. aprecruit.ucsf.edu/JPF05424

aprecruit.ucsf.edu

Junior/Assistant/Associate/Full Specialist Positions Available

University of California, San Francisco is hiring. Apply now!

02418

Reposted by Erik Thiede

Sam Blau @samblau.bsky.social · 14/05/2025

The Open Molecules 2025 dataset is out! With >100M gold-standard ωB97M-V/def2-TZVPD calcs of biomolecules, electrolytes, metal complexes, and small molecules, OMol is by far the largest, most diverse, and highest quality molecular DFT dataset for training MLIPs ever made 1/N

54610

Erik Thiede @erikhthiede.bsky.social · 08/05/2025

Yes, Bogdan is a smart guy who does great work: we met briefly at a conference two (?) years ago, and I've been following his work since.

020

Erik Thiede @erikhthiede.bsky.social · 08/05/2025

No worries, I'll start by reading that code. We are working on some techniques for cryo-EM that look mathematically similar to Lucy-Richardson, so I'm on the lookout for new tricks :)

110

Erik Thiede @erikhthiede.bsky.social · 07/05/2025

Can we get a reference for the algorithm? (I saw your bioRxiv preprint from last week, but am not 100% if that is the right paper to refer to.)

330

Reposted by Erik Thiede

Kim Weeden @weedenkim.bsky.social · 07/05/2025

One of grants frozen at Cornell: $6.7M (4 yrs) to develop heart pumps for babies born w/ heart defects. It took decades of work to get to point of prepping device for in-human clinical trials. No private company could do this. The cruelty of the Trump administration is heartbreaking. So to speak.

news.cornell.edu

Research at risk: Life-saving heart pumps for babies | Cornell Chronicle

After receiving a stop-work order from the federal government, the future of a device to help children with heart defects is uncertain.

8933611763

Erik Thiede @erikhthiede.bsky.social · 16/04/2025

Agreed: Universities are far too quick to ignore the words of the actual people in power, e.g. J.D. Vance saying, "Professors are the enemy" and "[Orban's seizure of universities is] the closest that conservatives have ever gotten to successfully dealing with leftwing domination of universities"

010

Erik Thiede @erikhthiede.bsky.social · 14/04/2025

Welcome to upstate NY: it's great up here!

020

Erik Thiede @erikhthiede.bsky.social · 01/04/2025

Our work also solves generalizability. As we know from a plethora of JCP, JPC, and JCTC articles, if something works on the alanine dipeptide it basically works on everything.

010

Erik Thiede @erikhthiede.bsky.social · 01/04/2025

This makes the latent encoding immediately interpretable. (Although paradoxically, as anyone who has taught free energy in an undergrad p-chem class knows, this doesn't mean it is explainable).

100

Erik Thiede @erikhthiede.bsky.social · 01/04/2025

The variational alanine encoder takes arbitrary data and transforms it into a latent random variable whose negative log-probability is given by the free energy surface for the alanine dipeptide in vacuum.

100

Erik Thiede @erikhthiede.bsky.social · 01/04/2025

It is with great pleasure that I announce that the Thiede lab has solved the problem of interpretability in AI for Chemistry with our new neural network architecture: the Variational Alanine Encoder (VAE).

160