We need to be preparing for this type of “weird” and “alien” reasoning applied to medicine. It will open up new opportunities for caring for people.
This post nails why we why we should take the Riemann Hypothesis breakthrough very seriously, even beyond math. We’re entering an era where solving problems will look very different from how we solved them in the past.
“This is not theory-building in the usual sense. It is something much weirder and more alien. The best human mathematicians know how to build theory, painstakingly, to reach a goal. But no human brain, on its own, is capable of creating a wormhole that requires a 25,000 page long tour through existing infrastructure with no obvious or realistic connection to the goal. But huge credit to Levent here: I think no current AI, on its own, is capable of this either.”
It used to be the case that to solve a particular problem - code, life sciences, math, cyber, engineering, and so on - you were ultimately bound by how much information an individual or team could retain and understand about the space. More information or more team members don’t always help, because it gets harder to connect the dots or coordinate amongst everyone. AI changes that completely.
For the first time ever we can decouple the human context limits and ability to work through information from our ability to drive productivity or discoveries. This is likely not just a small increase in the kinds of discoveries we can drive, but a qualitative shift in human output.
Now every single form of prior art, insight, past research, past decision, or unrelated breakthrough, can be used for the discovery or creation of something new. This is useful for anything from cybersecurity to life sciences research. Agents will be able to correlate entirely unrelated work to a separate project; scan every bit of data in your enterprise for solving a problem; and so on.
The solutions may not be as elegant as they were in the past, but we sure will have far more of them. We’re only in the first couple of months of actually understanding the implications of this but it’s clearly going to drive a lot of change.
2
187
Seth Hain retweeted
This post nails why we why we should take the Riemann Hypothesis breakthrough very seriously, even beyond math. We’re entering an era where solving problems will look very different from how we solved them in the past.
“This is not theory-building in the usual sense. It is something much weirder and more alien. The best human mathematicians know how to build theory, painstakingly, to reach a goal. But no human brain, on its own, is capable of creating a wormhole that requires a 25,000 page long tour through existing infrastructure with no obvious or realistic connection to the goal. But huge credit to Levent here: I think no current AI, on its own, is capable of this either.”
It used to be the case that to solve a particular problem - code, life sciences, math, cyber, engineering, and so on - you were ultimately bound by how much information an individual or team could retain and understand about the space. More information or more team members don’t always help, because it gets harder to connect the dots or coordinate amongst everyone. AI changes that completely.
For the first time ever we can decouple the human context limits and ability to work through information from our ability to drive productivity or discoveries. This is likely not just a small increase in the kinds of discoveries we can drive, but a qualitative shift in human output.
Now every single form of prior art, insight, past research, past decision, or unrelated breakthrough, can be used for the discovery or creation of something new. This is useful for anything from cybersecurity to life sciences research. Agents will be able to correlate entirely unrelated work to a separate project; scan every bit of data in your enterprise for solving a problem; and so on.
The solutions may not be as elegant as they were in the past, but we sure will have far more of them. We’re only in the first couple of months of actually understanding the implications of this but it’s clearly going to drive a lot of change.
Like most mathematicians, I believed the Riemann Hypothesis to be completely out of reach at present (even with the best AI tools), because I expected it to require substantial theory-building, probably in arithmetic geometry. The kind of new theory-building that fills entire books, and that AI isn't so good at yet.
Mathematicians have a notion of how deep a problem is. A sufficiently deep problem is generally intractable until enough of the theory is in place. If difficult problems are like valuable mineral reserves buried miles underground, you first need to build the infrastructure to reach them.
What Levent's proof seems to have done instead is something completely different. It takes an incredibly complicated tour through the infrastructure built for a hundred other mineral reserves, situated far from the destination, and somehow create a wormhole to the original destination.
This is not theory-building in the usual sense. It is something much weirder and more alien. The best human mathematicians know how to build theory, painstakingly, to reach a goal. But no human brain, on its own, is capable of creating a wormhole that requires a 25,000 page long tour through existing infrastructure with no obvious or realistic connection to the goal. But huge credit to Levent here: I think no current AI, on its own, is capable of this either.
I had never worked on or thought too much about the Riemann Hypothesis: it seemed very far removed from my own area of number theory.
But the astonishing thing about Levent's claimed proof of RH (which I think has a very good chance of being correct) is that it uses a huge number of precisely the things that are central to my own corner of number theory:
- Theta functions and other automorphic forms on unitary groups
- Gan–Gross–Prasad period formulas
- Relative trace formulas
Etc., etc.
A problem that I had always assumed would require new theories in arithmetic geometry may instead turn out to be tractable through an inhuman chain of existing (deep) results in higher-rank automorphic forms (and a lot else).
This morning I walk into my office at Epic for the last time. The future holds plenty of intrigue. Today, though, I'm grateful — 21 years' worth — for the customers who shared their dreams and challenges, the industry friends united in helping patients and caregivers, and the Epic colleagues who invented, sharpened one another's ideas, and relentlessly built.
You all taught me so much. Thank you.
Keep doing good!
Recently, a friend asked me, over a cup of coffee, what has AI actually changed around me personally. I immediately noted that the quality of understanding one can develop of topics is deeper through dialogue - and AI makes dialogue more accessible.
The combination of deep reading paired with dialogue - both with fellow humans and AI - is a skill we need to foster and teach.
1
11
4,930
Researchers at @Yale using Curiosity—Epic's generative medical event model, which simulates a patient's likely future course from their record — looked at adults discharged from the ED with abdominal pain, one of the most common and most uncertain presentations in emergency medicine. Traditional risk models answer a binary question: will this patient come back? Curiosity answers a richer one: if they return, when, and will that return end in discharge or admission?
Two patients can carry the same overall revisit probability but face very different paths—one headed toward a quick outpatient follow-up, another toward an early admission. A binary score flattens that difference. A trajectory surfaces it.
A single pretrained model, like Curiosity, can generate calibrated predictions across many outcomes without retraining. As a result, the focus and time shift away from building a bespoke model for every clinical question and toward deciding which questions are worth asking. The team moved from idea to results in a matter of weeks.
It also highlights an emerging skillset: clinicians who pair deep clinical reasoning with real fluency in generative AI. The questions they choose to ask are the key ingredient.
Preprint here: medrxiv.org/content/10.64898…
Find more on Curiosity here: epic.com/epic/post/curiosity…"
How does 5 years of data sound?
BMI and GLP-1 trends are now available on the Epic Research Data Tracker:
Longtime listener; first time caller. Had a blast exploring a wide range of topics - from scaling laws to Paperboy (the Nintendo game) - with @arjunmanrai and @AndrewLBeam .
In the latest episode of AI Grand Rounds, @sethHain, senior VP of R&D at Epic (@HeyEpic), describes how his company is building foundation models that respect institutional autonomy, minimize burden, and prioritize safety. Full episode: nejm.ai/ep39
Imagining a thoughtfully curated, hard-bound book (a la @stripepress) collecting the essential essays and papers demonstrating how society dealt with this moment in AI. Trying to compile what truly deserves to make the cut.
1
2
759
Framing to consider for the year(s) to come:
xitter.gq/ivanhzhao/status/20031…
1
1
390
Health systems are leading the way on AI research and adoption. It was an honor to share their real-world results with the tech community at #MSIgnite.
The speed and scale of what’s happening in healthcare right now is remarkable. And what’s coming next is even more exciting.
At Epic, we create technology to help people get and stay well.
AI is embedded throughout our applications—from the clinic, to the bedside, to the back office.
Worldwide, hundreds of thousands of clinicians are using it to improve care.
Here it is in action at #MSIgnite:🧵
Seth Hain retweeted
US News Best Leaders 2025: Judy Faulkner - “You predict the future for two reasons,” Faulkner says. “One is to prepare for it. And two is to change it.” usnews.com/news/leaders/arti…
Seth Hain retweeted
At @ChristHospital, 141 people started treatment earlier because Epic’s AI helped spot lung cancer in early stages.
Here's Mark's story:
Geniunely enjoy ever conversation I get to have with @mattlungrenMD and @JustinNordenMD - including this one we decided to record.
youtube.com/watch?v=k3q3xtoT…
Seth Hain retweeted
We really have not made a lot of progress on explaining the deep mystery of LLMs:
How does a model using matrix multiplication to predict the next word manage to simulate human thought well enough to do all the very human-like things it does? And what does that mean about us?
Seth Hain retweeted
Cool new paper out of Epic that trains foundation models + derives scaling laws for generative medical event modeling.
Interestingly enough - they find that the Chinchilla scaling parameters basically hold in this setting as well!
🧵some observations and questions below
Seth Hain retweeted
He was 21 when he programmed MyChart. Now he’s the president of Epic statnews.com/2025/08/22/ehr-… via @statnews @brittanytrang
☄️☄️☄️ Pleased to share our latest research on creating personalized medicine at scale by combining Epic Cosmos and generative AI.
We created CoMET, a decoder‑only transformer that generates next medical events to simulate patient trajectories. Trained on 115B medical events from Epic Cosmos; evaluated on 78 real-world tasks spanning diagnosis prediction, disease prognosis, and healthcare operations; established scaling laws on medical event transformers; and, importantly, determined that clinical performance improves along with scaling.
Grateful to both the Epic Cosmos community for making this possible and the collaboration across @Yale, @MSFTResearch, and @HeyEpic.
#ugm2025
arxiv.org/abs/2508.12104




















