steinly0.bsky.social

Reminder! arXiv is migrating our database to the cloud tomorrow (Wednesday, Oct 28th) starting at 7AM ET. From 7AM - 12PM ET arXiv will be read-only. More on the blog: https://blog.arxiv.org/2026/10/21/scheduled-database-maintenance/

An internal model at OpenAI has produced a claimed refutation of ETH! Russell Impagliazzo, Mohan Paturi, and I have been asked to help organize an open and independent examination of this claim. 1/4

“I almost felt a sense of relief.” That was Steinn Sigurdsson’s reaction to hearing about Valency. In his view, only AI can handle AI output at scale. Here’s what sparked it, and why he’s helping build the AI tools science needs: blog.valency.io/posts/adviso...

Fernando Pérez and Joshua Bloom always believed open-source software could transform science. Today it’s real, and Fernando is helping Valency bring AI tools to researchers. Read more on our blog: blog.valency.io/posts/adviso... #AIScience #Berkeley

🚨Attention🚨 Upcoming scheduled maintenance to arxiv.org. On Wednesday, October 28, arXiv will be migrating our database to the cloud, which will result in a brief disruption to arXiv service. Read more on the blog: https://blog.arxiv.org/2026/10/21/scheduled-database-maintenance/

AI always gives an answer, but only a domain expert can tell if it’s right. In astronomy that risks a flawed paper. In biopharma the stakes are far higher. Bond grounds AI in verifiable research so you can trust the results. Try MCP Bond free: valency.io/bond #AIScience #OpenScience #Berkeley

BREAKING: Major shakeup in NSF leadership. Brian Stone, who has been performing the duties of NSF director since April 2025, is retiring on Dec. 31. Rebecca Keiser, who has been acting chief of staff, is also retiring, Oct. 31. Unclear who will replace Stone, since O'Neill's nom hasn't moved.

How we’re approaching AI at bioRxiv and medRxiv Large language models and other AI tools are reshaping how research is done and shared. We’ve been listening to our community, and our new blog post shares our current thinking. Read it here → lnkd.in/gg_-JMhN

How we are approaching AI at openRxiv - new blog post prompted by growing adoption by researchers, submission patterns, and questions from the community openrxiv.org/approach-to-... 1/n

This Wed and Thu we’re joining @hf.co and the U.S. National Science Foundation in supporting the #OSIHackathon and its global network of researchers, students, and builders reshaping science through AI. Sign up to compete in our prize track for best use of MCP Bond 🏆 👉 luma.com/ku88xh92

All eyes on Tanegashima today for the 20:41 (⏰UK) #MMX launch to Mars' moon #Phobos! Getting an idea of scale for 🏴󠁧󠁢󠁳󠁣󠁴󠁿Radio: the spacecraft ≈ the size of a car, Phobos ≈ the size of the Isle of Arran, and the MMX rover ≈ the size of your carry-on suitcase. 🤩 🍿: www.youtube.com/watch?v=Fn6g...

Without the right guardrails AI can produce something that looks like science, but isn’t. It builds on shadows instead of the shoulders of giants. Bond keeps AI grounded in verifiable research: fewer hallucinations, results you can trust. Try MCP Bond free: valency.io/bond?utm_sou...

We’re excited to team up with Open Scientific Intelligence (OSI) to sponsor next week’s #OSIHackathon (Oct 21–22), including offering a prize track for the best use of MCP Bond 🏆 The event is free, hybrid, and global. 👉 Sign up today: luma.com/ku88xh92

As scientists ourselves, we recognize how important it is to build your own setup: tools, harnesses, workflows, and collaboration practices. That's why we've built MCP Bond to be LLM agnostic and integrate seamlessly with existing workflows. Try MCP Bond now: valency.io/bond?utm_sou...

GitHub gave open source a place to collaborate. Jupyter creator @fernandoperez.org says science needs its own: a hub where humans and agents interact but humans decide. Join the Hub waitlist: hub.valency.io?utm_source=bluesky&utm_medium=social&utm_campaign=2026-q4_hub-always-on&utm_content=co_bond

Pre-hoc: Volume…and trust. if noise:signal becomes vast the resource uses utility (and costs become unsustainable). Whatever one thinks of the new arXiv policy, that’s basically the reason. Post-hoc: trickier but we want parent institutions to take responsibility for outright fraud

Today marks an important milestone for the academic and scientific communities: OpenAI has released a broad range of new mathematical results. Valency is excited to host the collection on Valency Hub, open for all to read. Read the OpenAI collection: hub.valency.io/collections/...

I'm sure there's a good reason this door has two handles. #Iceland #photography #streetphotography #door #red

Swipe to see how Valency Bond keeps your AI current → Most LLMs are working off research that's months old. Valency Bond gives them work published hours ago. Save this for your next review, and get free access at valency.io/bond #AIresearch #LLM #research #science #preprints

🧪⚛️ Some thoughts on the past and future of the arXiv, a genuinely remarkable repository for human knowledge. nanoscale.blogspot.com/2026/10/the-...

arXiv imposes rate limit on paper submissions to stem the AI slop tide

As arXiv announces a new submission rate limit policy, I refer anyone who is up in arms about it to👇Hogg 2026👇. I'm so confused who would want to submit so much slop? For me (both writing and hiring) it's always been about quality over quantity, every single time. 🧪⚛️ blog.arxiv.org/2026/10/01/u...

The arXiv limit Putting my community service hat on, I thought I would relay an announcement on the an announcement on the arXiv blog that it is introducing a limit on the rate that any individual can submit papers to arXiv. To start with, the limit will be two per month. This limit applies to the…

arXiv rate limiting submissions in face of AI deluge blog.arxiv.org/2026/10/01/u...

The LLM-induced jump in preprints is most dramatic in computer science, but math is pretty clearly spiking as well, especially since the beginning of this year. info.arxiv.org/about/report...

I’ve known about this for a couple of weeks, but now I can share: as of today, arXiv is rate-limiting submissions to two per month. And as a mod, I have to admit this is a necessary move (at least temporarily). blog.arxiv.org/2026/10/01/u...

Attention arXiv authors. Because of the exponential increase in submissions, arXiv is limiting the number of papers that a person can submit each month to two. blog.arxiv.org/2025/10/31/a...

Your AI is answering with research from months ago. Valency Bond fixes that. It pulls new papers and preprints from PubMed, arXiv, bioRxiv, and more, and hands them to your LLM of choice. You get answers grounded in real sources, with citations attached. Get free access valency.io/bond?utm_sou...

This is important and high impact

It's going to take AI to keep up with AI. Valency is building the tools to get ahead of that and put it to work for the good of everyone. Sign up for Valency Bond for free: app.valency.io/start #AIScience #Berkeley

"If this goes on..." arxiv.org/stats/monthl...

New repository for math designed for the AI era. A key differentiator is apparently that authors aren't expected necessarily to understand/have performed the work in the way arXiv, etc would require. 1/2 hexagonmath.org

as I was saying arxiv.org/stats/monthl...

When September ends... arxiv.org/stats/monthl...

The last six months have opened up a whole new world for science. Anthropic, OpenAI, and Google are all shipping tools for research. Valency's mission is to meet scientists wherever they are, whatever model they're using. Sign up for Valency Bond for free: app.valency.io/start

arXiv has received a 17.2 million investment from three leading philanthropic organizations. Thank you to Simons Foundation International, Siegel Family Endowment, and XTX Markets for your support of arXiv! Read more: blog.arxiv.org/2026/09/23/arxiv-receives-multiyear-investment/

What does the AI-accelerated science era actually need? Not just tools for astronomers or STEM, but a home for every researcher trying to engage with AI in an authentic way. That's what Valency is about. Sign up for Valency Bond: app.valency.io/start (free for Academic Researchers)

One problem with AI: it always gives you an answer. And as a scientist, always getting an answer should worry you a little. You still need people to look at the output and say, "I don't believe it," then dig into why. Sign up for Valency Bond: app.valency.io/start (free for Academic Researchers)

The exciting moment for AI as a tool will be when it recedes into the background, still everywhere, just no longer worth mentioning, the way we stopped talking about CCDs and big telescopes. Sign up for Valency Bond: app.valency.io/start (free for Academic Researchers)

AI today is fantastic at answering questions we've asked before. The frontier now is whether it can hypothesize beyond what we already know. Not a replacement for scientists, but a co-thinking machine to accelerate work. Sign up for Valency Bond: app.valency.io/start (free for Academic Researchers)

Could Planet 9 already be sitting in our databases, waiting for the right algorithm to find it? AI could help astronomers answer that, along with what's in the atmospheres of other planets and whether there's evidence of life. Sign up for Valency Bond: app.valency.io/start

Meet Fernando Perez (@fernandoperez.org), associate professor of statistics at UC Berkeley's College of Computing, Data Science, and Society and faculty scientist at Lawrence Berkeley National Lab. Read more about him here: blog.valency.io/posts/adviso...

this

With a Better Understanding of Physics, We Could Predict Volcanic Eruptions www.wired.com/story/physic...

Fields Medalists: "how to make sure that, as AI changes the way work is done, we don't lose sight of what that work was meant to achieve...whether these changes benefit the field or have a destructive effect will in large part be determined by the decisions of the humans in control" mathandai.org

✨NEW: bioRxiv and medRxiv now let you add or update funder information after your preprint is posted. No need to create a new version. Just update in your Author Area. Rolling out now with updates live within 72 hours. Check out the blog to learn more → openrxiv.org/updating-fun...

Google Scholar launches "Quick read". Will this be a useful filter IDing papers one actually wants to read (rather than having to skim then bounce out of those that clearly aren't) or might it discourage even more drive-by citations? scholar.googleblog.com/2026/08/quic...