Ben Blaiszik

What if We Just Built It All? A Call to Action to Reimagine Scientific Software in the Age of LLMs Scientific software is the invisible infrastructure of modern research, and it's been starved of software engineering resources for decades. With LLMs and coding agents, that's changing.

Cloudflare rebuilt Next.js in a week with AI for $1,100. I wrote about how this could become a defining moment for the scientific community to erase decades of tech debt and build the software tools that we have imagined, to accelerate progress. Let's go!

www.linkedin.com/pulse/what-w...

I was looking at my dissertation defense slides recently and was fondly like... "Wow, this is what it's like to focus on one project a year."

The second panel will feature researchers from industry including Santiago Miret (Lila), Abhijeet Gangan (Periodic), Lily Kim (Microsoft Discovery), and Tim Erdmann (IBM)

More details on the panel here, with registration: www.marda-alliance.org/2026-marda-a...

The Materials Research Data Alliance meeting is next week! Great panelists and opportunity for connection. Registration is free, and content is available online worldwide.

Sign up to hear from key speakers: James Warren (NIST), John Schlueter (NSF), Andrew Schwartz (DOE), and Sean Donegan (AFRL)

Portfolio: fengzhihao77.github.io

If your team works on ML for materials/molecular discovery and needs someone who has built physics-informed networks for real problems, check out Zhihao Feng. Genuine problem-solver, strong collaborator, adaptable. Here's an example of his work.

Link: www.linkedin.com/posts/fengzh...

What are the biggest software-solvable problems holding you back in your research?

e.g.,
- There's no Python package to convert formats between X and Y
- Running tasks on HPC systems requires expertise I don't have
- Formatting or tracking the references for my paper is annoying

As a kid, I was outdoors and in the woods all the time. The pandemic reminded me how awesome outside is. I still do as many meetings as possible on walks.

or you'll just publish...much more. 😅

"If I had a million dollars
We wouldn't have to eat Kraft Dinner
But we would eat Kraft Dinner
Of course we would, we'd just eat more"

The 29th is a typo on the final day, it is in fact the 19th :)

There will also be opportunities to connect with others interested in these topics through poster sessions and a community Slack.

The event is online, so available around the world. Full speaker list soon!

The 4th Materials Research Data Alliance meeting is set for Feb 17-19 on topics in materials + data + AI. This is a great opportunity to hear from speakers from academia, national labs, and industry.

Register now (free - online): marda-alliance.org/2026-marda-a...

I'm excited to help improve our understanding of catalysis and fusion materials via the Integrated Scientific Agentic AI for Catalysis (ISAAC) and CascAIde projects; and to see Rick Stevens and @ianfoster leading ModCon to develop/collect the models/data needed to advance science. Stay tuned!

DOE announces $320M in new investments in the AI4Science!

You can read more about the Genesis Mission, the American Science Cloud, the Transformational AI Models Consortium (ModCon), along with 51 AI in science application projects here and in the attached image.

🔗 energy.gov/articles/ene...

Amazing week of results from AI4Science with NeurIPS and MRS. But, for those who couldn't make it, post a link to your research here we can boost you too!

I saved this image from Stable Diffusion in Dec 2022 because it was amazing a model could output such detail. 😅

Release v0.4.1 · TorchSim/torch-sim Thank you to everyone who contributed to this release! This release includes important bug fixes, new features, and API improvements. @thomasloux, @curtischong, @CompRhys, @orionarcher, @WillEngler...

Full details and package here. github.com/TorchSim/tor...

Paper: arxiv.org/abs/2508.06628

Thanks to the 10 contributors! Thomas Loux, Curtis Chong, Rhys Goodall, Orion Archer Cohen, Will Engler, Abhijeet Gangan, Andrew_S_Rosen and others

Bonus, a fun NanoBanana interpretation of the TorchSim paper

New features include mixed periodic boundary condition (PBC) support, NVT Nose Hoover method, and more. This release also enables important integrations with QUACC and Atomate2 to proceed. More on that soon!

TorchSim continues its growth aiming to be the high-performance engine for MLIP-powered atomistic simulation.

With this release (4.1), there improvements across the entire stack with new features, bug fixes, and improved documentation.

So your data are available upon reasonable request? Well, we are making some reasonable requests - at scale. :)

1. Search literature (currently stubbed)
2. Enumerate papers, extract contacts
3. Send email w/ data drop location
4. Parse data

Does anyone want to help productionize this?

Check out our updated website, publish your data, or leverage the 600 TB of data we have on hand at
www.materialsdatafacility.org

- First, a new section showing how to use MDF's Foundry datasets. Load ML-ready datasets with just a few lines of code
- Second, a clean hero, with updated Navbar and clear Calls-to-Action
- Third, a more compact and intuitive view of Foundry dataset contents, browsable from the search interface

Within Cursor, Gemini 3, and Claude missed simple things that caused component errors. Sonnet fixed these immediately in CC, showing the importance of the harness.

There were lots of little updates, but the attached images show a few specifically

2. The model harness matters...a LOT. That means that you might see vastly different performance e.g., in Cursor vs Antigravity (Google's IDE) vs Droid.
3. The cost was low. This update took 2h, $0.90 Cursor credits, 2 free Antigravity prompts and 1 Claude Code (CC) prompt.

Yesterday I did a glow-up of the entire Materials Data Facility front end using Gemini 3 in less than 2 hours. A few things I learned followed by few snapshots of the updates.

What did I learn?
1. Gemini 3 feels clearly ahead in terms of front-end design. Less AI slop, and more focus.

Northern Lights over the western Chicago suburbs at 9:15. Clearly visible by eye, lasted about 10 mins. 🤩

So glad you are getting some relief from all of this! 💜

Did a quick experiment last night to see how far kids run on Halloween by giving my son a phone. The answer was 5.3 miles.