🔥Telegram İfşa

Love this research direction. When you switch back and forth between running MD and training neural networks it’s hard to escape the sense that they’re really doing the same thing at some fundamental level.

Introducing CECAM-US-WEST, a new CECAM node located at UC Berkeley. Programing starting next calendar year! College of Chemistry CECAM chemistry.berkeley.edu/news/new-inter…

cqt.sg/highlight/2025… If this makes you laugh (at us) and then makes you think (about fluctuation theorems), consider nominating us for the igNobel prize :)

Tell me your company is busy building out its tech, without telling me your company is busy building its tech. normalcomputing.com

Postdoc opportunity! Join us in heavenly Vancouver (Canada) to develop fundamental nonequilibrium stat mech, thermo, and info theory applied to biomolecular machines and in close collaboration with experiment. Details: sfu.ca/physics/sivakg…

What goes around comes around. Back in the dot com heyday there was a startup that sent their employees to SF union square to hand out flyers wrapped around $2 bills. I still have the $2 bill. I have no recollection what the startup did.

Fun new paper by Callum SFU Physics, on energetically efficient multi-parameter driving of our favorite molecular motor (shout out F1!). doi.org/10.1103/bm3d-c… Supported by NSERC / CRSNG TIPS_SPIIE

The principle of covariance and fluctuation theorems are combined into a coherent framework of relativistic thermodynamics, applicable to moving thermodynamic systems and moving heat baths go.aps.org/44kDHwk

New preprint out today: “Information Thermodynamics of Cellular Ion Pumps”. Led by talented undergraduate student Julián Jiménez-Paz, and working with David Sivak, we explore the thermodynamics of cellular ion pumps like the sodium-potassium pump.

Absolutely well-deserved honor! Congratulations to Professor Amari and everybody working in related (huge) fields!!

So we opened the world’s first Chemifarm today! chemify.io

Normal EDA is a purpose-built AI platform to solve the semiconductor industry's most pressing challenges. This is the future of verification: consistent, objective, and auditable. Built for the complexity of modern chips and designed to scale as silicon systems grow more

Still accepting postdoc applications!

Interesting paper by Stephen Whitelam at LBNL that builds on our work on thermodynamic computing. In 2023, our team at Normal proposed to implement diffusion models on stochastic hardware, in our paper "Thermodynamic AI and the fluctuation frontier". Here, Stephen investigates

VerilogEval is an odd coding dataset - it gives the agent access to the golden testbench, which you will use to evaluate the final design from the model. I'm honestly surprised how papers like VerilogCoder or ChipAgents only get ~95% accuracy in this setting... That's what

Every smart person I know is getting into hardware right now

1/ Where do Probabilistic Models, Sampling, Deep Learning, and Natural Sciences meet? 🤔 The workshop we’re organizing at #NeurIPS2025! 📢 FPI@NeurIPS 2025: Frontiers in Probabilistic Inference – Learning meets Sampling Learn more and submit → fpiworkshop.org

5/ 📢 Don’t miss this multidisciplinary lineup of speakers: • Gavin Crooks — Gavin Crooks • Arnaud Doucet — Arnaud Doucet • Maxim Raginsky — @mraginsky • Ellen Zhong — Ellen Zhong • Pratyush Tiwary — @tiwarylab • Rianne van den Berg — Rianne van den Berg

Really excited to announce our new preprint, "Solving the compute crisis with physics-based ASICs!" For years, I’ve argued we are returning to the era of specialized hardware. We moved from CPUs to GPUs, then to domain-specific accelerators. We are now facing a clear compute

5 years after Sara Hooker ‘s hardware lottery paper, it’s great to see a map of what the new and meaningful compute building blocks can look like. arxiv.org/abs/2009.06489