Eliot Eshelman 🤖🔭 (@hpc_twit) header
Eliot Eshelman 🤖🔭 (@hpc_twit) profile picture

Accelerating scientific AI platforms @NVIDIA. From muons to bacteriophages to astrophysics. Opinions my own. he/him #Inclusion

Boston, MA
Joined March 2014
Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)Photo by Eliot Eshelman 🤖🔭 (@hpc_twit)
Who’s building Claw@Home?
In my ongoing quest to find the limits of Opus 5.5, I directed it to audit my old scientific field, theoretical astrophysics. We indexed about 150,000 papers, audited 50,000 for issues, and did progressively more thorough checks on thousands more. We found nearly 1000 major errors that got cited many times, most are still uncorrected. This survey enabled me to estimate the efficacy of peer review for correcting errors (not very good) and to identify dozens of open questions that anyone with spare Claude Max tokens could answer today (.md context files included). Bottom line: AI can speed up quantitative science by >1000x with no further model improvement required. But I'm seeing no evidence of this being pushed within universities, only academics complaining about labs dropping math proofs. Among my fellow reformed academic high agency founders, however, several are speed running entire disciplines. This survey has implications for how science will be run in the very near future. Hint: Academic publishing is dead. It was already a very stupid way to record and promulgate knowledge, and had been so for decades, even before all that publicly funded knowledge was locked behind paywall cartels. Now AIs can generate thousands of novel, correct papers per day, it is no longer meaningful for humans to consider a research paper a "unit of science" or to even bother much with the 99.9% of science which, up to this point, was necessary drudgery. Time to discover something like General Relativity or Evolution every week. LFG!! Enjoy: science.caseyhandmer.com/phy…
1
46
Hopefully improving agent behavior is more tractable than improving human user behavior 😅
At Harvard, agents generate a ton of jobs and pound the file system far more than what humans do; eg 7.9x more tree walks and 5.9x more recursive grips than humans. Agents cancel batch jobs 5x more than humans too. Buckle up, because this is the future. #SC26
1
106
Eliot Eshelman 🤖🔭 retweeted
At Harvard, agents generate a ton of jobs and pound the file system far more than what humans do; eg 7.9x more tree walks and 5.9x more recursive grips than humans. Agents cancel batch jobs 5x more than humans too. Buckle up, because this is the future. #SC26
4
9
1
42
2,178
I love beautiful science
I asked Gemini Carbon in @antigravity to create an interactive playground of famous quantum physics experiments. You can fire electrons at double slits, watch photons tunnel through barriers, paint your own potentials, and watch single-particle hits build up live on a detector wall. For example in the video you can see the fringes in the double slit experiment, but not in the tunneling experiment. Under the hood, the app is solving the 2D Time-Dependent Schrödinger Equation in near real time in WebGL2, sampling each detector click directly from the live probability wave as it arrives.
3
245
Eliot Eshelman 🤖🔭 retweeted
🧬 We’re hiring at the NVIDIA × Lilly AI lab—across both companies! Help build the next generation of drug discovery, bringing AI, robotics and experimental science together in our joint Bay Area lab. #Hiring #AIforScience #DrugDiscovery
13
61
9
599
81,964
Everyone should transition from test-driven-development to eval-driven-development, right? Is there a way without descending into a Goodhart nightmare?
81
Eliot Eshelman 🤖🔭 retweeted
Amazing milestone before 1 year for @EdisonSci. Nov 2025, we launched a platform for individual scientists. - 2.5 months later, we on-boarded first enterprise - 7 months from launch, first public pharma deal with Incyte - 8 months, went live with first internal deployment - 11 months, first public top-10 pharma deal I know in tech, this feels achievable. In pharma, this is an insane trajectory. What a ride so far!
12
11
133
5,119
I also used Claude Code to create three separate home networks: one for guests, one for IoT devices, and one for our trusted devices. That being said, I don't think my grandmother could have done it. As smart as they are, these AI systems still require some baseline level of instruction on what their tasks and goals are.
# Black October For anyone who knows me even somewhat well, it's clear that I am a 99th percentile tech optimist. However, the vibes have shifted for me quite a bit in the last 18 hours, and I feel a moral imperative to share what I'm thinking and feeling with my circles so that everyone can understand, and hopefully take seriously, what's happening right now. I think October 2026 will have many chapters written about it in the history books. You know those scenes in financial movies like Margin Call where someone realizes that they're all fucked, and no one else sees it yet? And that person feels like they're crazy, and then suddenly everyone is pulling an all-nighter in a boardroom and freaking out? Yeah, that's kind of how this is feeling right now in the SF/Bay Area/AI circles. We're 48 hours into what is being called Black October. In the last 2 days there have been several dozen cyber attacks which, in isolation, would each have been the worst to happen in an entire calendar year. Every sector from government to education to big tech and banks is being massively affected, and people are starting to freak out. In a nutshell, what's happened is that frontier AI models have gotten to a point where they are as advanced as the most sophisticated programmers, with one critical difference: they can work 24/7, and you can spin up 10,000 of them with a single command and they'll work together tirelessly until they accomplish their task. The most advanced models today are from companies like Anthropic and OpenAI, whose current best technology is so good that they haven't been able to roll it out to everyone yet because of the national security risks it poses. Meanwhile, the best open source models (where anyone can just download the model weights and do whatever they want with them, uncensored) are slightly behind the frontier but still 90-95% as capable. These models, developed primarily in China, are being used by black hat hackers to wreak havoc on the Western capitalist structure. Presumably, this has been part of China's plan the entire time, and we're seeing its implications fall out in real time. In red-pilled AI circles, we've been sharing and reading papers called things like "Situational Awareness" and "AI 2027" for the last year or so, which at the time felt a bit like near-term speculative fiction. They laid out the case for why Artificial General Intelligence (AI smarter than any human at any knowledge task) and Recursive Self-Improvement (models that are able to use themselves to get better) weren't half-baked future ideas anymore, but instead an inevitable outcome of the linear progress that's happening. These papers would then talk about things like "takeoff scenarios," trying to envision how that moment would come to be and what it would be like when it happened. It would either be a "slow takeoff" that would play out over the course of months and quarters, happening subtly and only obvious to a few. Or, it would be a "fast takeoff" where it would pretty much happen instantly, like a bit flipping from 0 to 1. From everything that I can see and am feeling, we are currently experiencing the fast takeoff scenario. Beyond the crazy security vulnerabilities that have come out recently, there have also been massive breakthroughs in life sciences, fundamental computer science, materials science, and pure mathematics that have all landed in the same time frame. While I still firmly believe that AI will be a net positive for humanity, it's more clear to me than ever that the societal and individual risk is very real, and is playing out NOW. For as much conviction as I have that this is happening, the thing that's caused me to lose sleep is that I don't really know what to do about it, or more importantly, what to tell you, reader, to do about it. So, the best I can do is make a few suggestions that have always been a good idea but are more timely now than ever: Set up 2-factor authentication for everything. I use Duo. Use a password manager across your devices. I use 1Password. Get an identity protection app set up. I use Alta. Move your IoT devices to their own network. Claude helped me do this. And last, but certainly not least, if you've been avoiding leaning into AI, see this as a catalyst to wake up and start using it. Don't just use the chat mode version of ChatGPT/Claude; click that Code/Work button and see how easily you can have AI write code to help you with your job. One of the biggest vulnerabilities was in a massively popular programming language called PHP, and the prompt that set it off was literally, "Find vulnerabilities in php." This is an unprecedented time for humanity, but as we've done with many things before, we will prevail so long as we are kind, loving, and open-minded to the world. Much love.
2
100
Great talk! Gives you a sense of how to ride continued AI progress. Also gives specific advice on what you can do today.
Here is the full talk:
1
121
The singularity asks you to wait. Whether you expect paradise or catastrophe, the story is the same: history is converging on a point, and what you do won't matter much. Nietzsche saw this pattern 150 years ago. When people believed history was a rational process heading toward its conclusion, he watched them become spectators, either resigned latecomers or worshippers of whatever was winning. He called it a historical sickness. The singularity revives that sickness in a new form. Some people are paralyzed: why build, learn, or create if machines will soon do it better? Others treat it as destiny and stop asking who is steering, or toward what. Both reactions hand over the present. But the future isn't a place we arrive at. It's made, decision by decision, by people who keep acting as though their choices count, because they do. Take the forecasts seriously. Prepare. But don't mistake a trajectory for a fate. The question isn't whether the singularity is coming. It's whether you'll be living, working, and choosing in the meantime, or waiting.
57
Imagine a podcast being released just as the printing press was hitting the world: what would world leaders and religious authority figures have to say? This episode has a little of this flavor and will get you thinking. buff.ly/t6zwvkx
44
Eliot Eshelman 🤖🔭 retweeted
What’s going on with these results, is the universe having floating point errors??
You can do Fourier transforms in O(N log(N)^0.9999999999999) time. Fuck this gay chud universe
41
48
17
3,680
219,660
Eliot Eshelman 🤖🔭 retweeted
A universal predictive model of the cell would dramatically accelerate science, allowing biologists to perform experiments digitally. Today we’re announcing a major partnership with the DOE and NIH. The goal of this partnership is to generate the data that will be needed to build an accurate model of the cell with artificial intelligence. Isomorphic, Google DeepMind, and Meta are joining us as founding partners in the Virtual Biology Initiative. This is the beginning of a large scale coordinated effort to map cellular biology to power the development of digital models of life. Other leading scientific institutions and consortia including the Allen Institute, Broad Institute, Gladstone Institutes, the Human Cell Atlas, the Human Protein Atlas, the Wellcome Sanger Institute, as well as NVIDIA, are working together as part of this international scientific project. Creating a predictive model of the cell is one of the most important challenges for the next decade of science. The insights that come from this could unlock a far greater understanding of disease, and open new paths for cures. We invite the worldwide scientific community and other funders to join us in this effort.
106
247
102
1,778
264,374
Eliot Eshelman 🤖🔭 retweeted
AI can increasingly make progress on open problems like Navier-Stokes. Yet defining a new problem still relies on human research taste. Introducing ScholarCatalyst, a benchmark built from AI researchers’ firsthand accounts of what inspired their work.
11
65
17
301
66,084
Eliot Eshelman 🤖🔭 retweeted
Schmidt Sciences is hiring Fellows in Residence! You help us with grantmaking decisions, while continuing your own research - $50k compute, $50k for a workshop - based in NYC - past fellows kept affiliations with Stanford, MIT, JHU Apply by Dec 15 Link 👇 Please share/rt :)
3
43
234
28,875
Eliot Eshelman 🤖🔭 retweeted
Join us at the MIT Media Lab for the ScienceClaw Hackathon, building the internet of agents for science. AI is becoming a collaborator in the real world - designing materials, creating instruments, running experiments, and connecting its capabilities with those of other agents. AI adapts as a problem unfolds - revising its reasoning, learning how to collaborate, and assembling scattered pieces of knowledge, evidence, and raw capability into solutions to some of the hardest challenges in science, technology, and innovation. Teams will connect AI agents, models, simulations, robots, cloud labs, and scientific tools into functioning systems to tackle problems in protein design, robotics, materials, manufacturing, experimental science, and beyond. The challenge is to turn ideas into tested results - and demonstrate how agents collaborating across teams and disciplines can accomplish more together. Explore how agents can specialize, challenge one another’s assumptions, learn from failed experiments, and combine their expertise to solve harder problems. Show how collaboration changes what your system can discover, design, or build. One agent's discovery becomes another's starting point. A tool built by one team enables an experiment by another. Connect your team of AIs, share a capability someone else needs, and build on what others have learned, produced, built. The ambition is an internet of agents through which scientific knowledge, tools, and capabilities can grow, evolve and be utilized across teams and institutions. 🗓️ When: October 30-November 1, 2026 📍 Where: MIT Media Lab Bring your expertise, your tools, and a problem worth solving - or find one. Help build AI that can contribute to science through what it can discover, create, and make work.
19
36
7
198
36,543
Awesome let’s get cracking!
When I read that an internal OpenAI model solved the Navier-Stokes problem, I realized it would be very hard for us to know if that model possesses similar "millennium intelligence" in biology (even if it does!). "Curing disease" is a noble goal, but we cannot validate proposed treatments rapidly enough to use it as a frontline eval for new models. Here, @SGRodriques and I propose Millennium Problems for Biology - meaningful, difficult challenges in biology whose solutions should be straightforward and quick to validate in a wetlab. If a model were to solve any of these, I would know that machine intelligence was capable of fundamentally reshaping humanity's ability to make, measure, and model biology. Please see details of the problems in the thread below, and see our website for the permanent home of the list. I encourage the community to red-team these challenges to make sure they are adequately specified and difficult to cheat, and propose additional challenges to add. We will subsequently update the list in response to feedback.
1
278
Eliot Eshelman 🤖🔭 retweeted
ToolUniverse just crossed 1M downloads on PyPI! 🎉 A huge thank-you to everyone who contributed tools and skills, reported bugs, shared ideas, or built something with it. You’ve helped turn ToolUniverse into a growing open ecosystem for AI scientists. And v1.5.0 is here: 🔬 2,700+ scientific tools 🧭 180+ research skills The new release brings Boltz for structure and binding prediction; dbGaP and EGA for controlled-access genomics; GIAB benchmarks, DANDI, the FDA Purple Book, Consensus and Exa search; and AlphaGenome Atlas support. One part I’m especially excited about: 30 new skills for turning heavyweight local models into tools your agent can actually call—including scVI, Harmony, Monocle3, Squidpy, CellRank, ESM, Enformer, ChromBPNet, MACS3, LDSC, and more. Each skill walks you through setup, validation, and troubleshooting, while your models and data stay on your own machine. From protein structure to single-cell analysis, your AI agent can now do much more—without sending the underlying data elsewhere. Install everything with one prompt: “Read aiscientist.tools/setup.md and set up ToolUniverse for me.” github.com/mims-harvard/Tool… aiscientist.tools/
2
1
13
929
We’re getting there! They need to be able to get their hands dirty, though. Will be interesting to see that work.
Today we’re releasing Helix 2.5 We rented 30 homes in the Bay Area. The robots arrived with no additional training and started doing useful work
1
186
Eliot Eshelman 🤖🔭 retweeted
Biology does not need models that just describe the world. It needs models that predict what happens when we intervene. In @CellCellPress Focus on AI in Biology, we lay out how to build biomedical world models, from the data they need to how they should be trained and evaluated.
10
81
8
452
26,249