Michael Bronstein

Otherworldly misty morning in Chisapani forest on a Marwari stallion in a dreamy photo that looks unreal. If equestrian paradise exists, it should look something like this.

@PhilipProudfoot Again, you are being very charitable. A missile costs at least half a million.

@mmbronstein Ah well, yeah, you’re right. Probably should execute him.

Israel attempting to murder a journalist, caught live on on camera.

@PhilipProudfoot Israel is succeeding at what Hegel, Schelling, and Fichte couldn't accomplish -- completing the system of German Idealism.

🚨 Iran is strategically holding back its doctors of philosophy. Larijani was a moderate Kantian. DC fears the escalation ladder: Hegelian → Schmittian → Heideggerian → post-structuralist At the final stage, the subject dissolves, the target disappears, power is everywhere

@BowesChay @SweeneySteve Russia Today is state propaganda.

My close friend and colleague Steve Sweeney @SweeneySteve is nearly killed by an Israeli strike in Lebanon. Their vehicles are clearly marked as Press. They wear Press markings. He's alive. What more can we say.

We recently taught a short course at the ENAR 2026 Spring Meeting on generative models for protein, cell, and biomedical data. We’re excited to share the course materials here for anyone interested: pengzhangzhi.github.io/ENAR26-Course-… with @Anru_Zhang, @AlexanderTong7

There's an old joke in systems biology called "How Biologists Fix a Radio." A biologist, tasked with figuring out why a radio doesn't work, removes components one by one and catalogs the result. Remove this transistor: the radio makes a horrible screeching sound. Conclusion: this is the "horrible screeching transistor." Remove another component: the radio goes silent. Conclusion: this is the "silence transistor." This is essentially what we do with genomics. We see which genes are mutated in cancer and assume they must be "cancer genes." We see which genes are differentially expressed and assume they must be "important." But correlation is not causation, and a parts list is not a circuit diagram. You can have a complete inventory of every resistor, capacitor, and transistor in a radio and still have no idea how it plays music.

i’ve been reflecting on how much my own behavior has shifted from reading scientific papers to asking AI to interpret the literature. as LLMs get better, more of our interaction with knowledge is mediated through generated summaries rather than primary sources. that shift is not just about how we consume information. it undermines the economic layer that has historically funded truth generation, across both journalism and science. in news, if LLMs can produce infinite “journalism-like” articles, the marginal value of content collapses. subscriptions erode, ads weaken. when anyone can generate something that looks like reporting, the institutions that fund actual reporting, including investigation, sourcing, and verification, start to break. the traditional business model was already cracking, and AI may collapse it entirely. in science, the same dynamic plays out. if LLMs can generate “paper-like” PDFs, the supply of plausible research explodes and signal is lost in the noise. journals and citations, already imperfect proxies for truth, become even less reliable. when publishing is cheap, it stops being a meaningful filter for correctness. the incentive shifts toward producing more papers, not more correct ones. the core issue is that our systems reward the production of content, not the generation of truth. journalists get paid to publish, not to be right. researchers are rewarded for output, filtered through peer review systems with no skin in the game. reviewers do not profit from identifying important work or lose from endorsing weak work. prediction markets offer a different architecture, one that shifts incentives from output to accuracy. instead of rewarding publication, markets reward correct forecasts. if you uncover a scoop, generate a dataset, or replicate a result, you can monetize that knowledge directly by taking a position in a market tied to the truth, then revealing the information. this changes the unit of value. it is no longer a paper or article, but a resolved question, such as whether a clinical trial succeeds or a result replicates. anyone who can answer these questions early, including journalists, researchers, labs, and AI agents, has an incentive to do the hard work of discovery and verification. this is particularly powerful for science. today, novelty is rewarded over correctness. replication is undervalued, and null results never get published. in a market system, the incentives flip such that shorting a flashy result or replicating an overlooked finding are profitable. the broader shift is that journals, news outlets, and preprint servers become oracles feeding into markets, rather than the primary locus of value capture. the economic reward flows to whoever is most accurate about reality before it is obvious. AI makes content cheap, which makes correctness more valuable. prediction markets may be one of the first mechanisms that directly reward it.

@RRathore334001 Once you ride one you get hooked

@mmbronstein how did you come to love marwaris?

Prithivi, my favorite Marwari horse

@POLITICOEurope x.com/mmbronstein/st…

Ex-Conservative MP Crispin Blunt was charged today with four drug offenses. politico.eu/article/former…

@TweeterReynolds @Conservatives x.com/mmbronstein/st…

Crispin Blunt was the only high profile @Conservatives MP who took a progressive, intelligent, evidence-based approach to #drugspolicy. He was also strongly opposed to Israel's occupation and subjugation of Palestine. This is his reward. bbc.co.uk/news/articles/…

@fionasins_ x.com/mmbronstein/st…

Sorry the MP named Crispin *Blunt* was charged with possession of weed??

We congratulate Gerd Faltings as the 2026 Abel Prize laureate! 🎉 He recives the Abel Prize "for introducing powerful tools in arithmetic geometry and resolving long-standing diophantine conjectures of Mordell and Lang".

@JohnSimpsonNews @RoleMarks One will sooner see a wooden horse pee, as the Russian expression goes.

@RoleMarks Of course.

The Geneva Conventions make it clear that attacks may only be directed against military objectives. A civilian object like a block of flats loses its protection only if it is being used to make an "effective contribution to military action."

@Never_Again2020 "This story shall the good man teach his son; And Crispin Blunt shall never go by, From this day to the ending of the world"

Former Tory MP Crispin Blunt, a vile antisemite who claimed that Jews “demand special status”, compared Palestine to the Holocaust and called for “eliminating subsidies” to the Community Security Trust (CST), has been charged with four drugs offences. bbc.co.uk/news/articles/…

🚀MIT Flow Matching and Diffusion Lecture 2026 Released (diffusion.csail.mit.edu)! We just released our new MIT 2026 course on flow matching and diffusion models! We teach the full stack of modern AI image, video, protein generators - theory and practice. We include: 📺 Videos: Step-by-step derivations. 📝 Notes: Mathematically self-contained lecture notes 💻 Coding: Hands-on exercises for every component We fully improved last years’ iteration and added new topics: latent spaces, diffusion transformers, building language models with discrete diffusion models. Everything is available here: diffusion.csail.mit.edu A huge thanks to Tommi Jaakkola for his support in making this class possible and Ashay Athalye (MIT SOUL) for the incredible production! Was fun to do this with @RShprints! #MachineLearning #GenerativeAI #MIT #DiffusionModels #AI

🚨 We’re Hiring in Generative AI for Biomolecular Design! 🧬🤖 We’re looking for Research Scientists working on biomolecular design, including small molecules, proteins, RNA, and molecular dynamics. Join us to help shape the next generation of AI-driven discovery, expanding our frameworks (e.g., Proteina-Complexa, La-Proteina, ReaSyn, and GenMol) to new domains and problems. If you’re excited about pushing the frontier of generative AI for biology and chemistry, we’d love to hear from you. 🔗 Apply: nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAEx… 🌐 Group page: research.nvidia.com/labs/genair/

@PhilipProudfoot @JohnSimpsonNews No, his first day as an activist

@JohnSimpsonNews Is this your first day as a journalist, John?

I don’t get it. If Hezbollah had flats in this block (unlikely in central Beirut but not impossible) why did the Israelis give everyone inside an hour to get out — including those they wanted to kill? And if there weren’t any Hezbollah people there, why destroy a building with dozens of civilians in it?

@Andr3yGR Yes

@mmbronstein Do you mean Penrose? I always thought of him as a mathematical physicist. Perhaps that score is not fully settled yet :)

A computer scientist takes the Physics Nobel. A physicist takes the Turing Award. Balance is restored ⚖️

📢📢 Proteina-Complexa 📢📢 Atomistic Binder Design with Generative Pretraining and Test-Time Compute + Experimental Validation at Scale ⭐️ Project page (research.nvidia.com/labs/genair/pr…) for: 📜 Method paper (ICLR 2026 Oral) 🧬 Wet lab paper 🛠️ Code & models 📁 Data 🧵 Thread (1/n)