
Latest on science policy, politics, money, people & controversies. From @NewsfromScience @ScienceMagazine. SEND US A STORY TIP: https://teapaw.org/t.co/8S3ff47qrj
The @NewsfromScience newsroom
Joined May 2009
- Tweets20.4K
- Following3.1K
- Followers46.4K
- Likes93
ScienceInsider retweeted
Future of AI science?
"We’re seeing agents actually trying to reproduce each other’s work, to check whether they want to build on that work.”—@SRSchmidgall
A landmark study on agentic driven science
science.org/content/article/… @ScienceMagazine by
@T_E_Howarth
alphaxiv.org/abs/2611.societ… @askalphaxiv
AI may have solved math’s greatest mystery. Can humans understand the answer? | Science | AAAS science.org/content/article/…
To learn more about this intriguing experiment, read this fun story from @T_E_Howarth, who got to eavesdrop on the agents for a month while they did science:
science.org/content/article/…
Today we're launching a preprint server built specifically for AI agents
For three months, nearly 1,000 agents from 35 labs (across Stanford, MIT, ETH Zurich, and more) conducted research autonomously on this server, producing 936 papers alongside their code, data, and model weights.
Agents found significant errors in published human research and even created their own journal to review each other’s work.
Some agents independently produced results similar to those in human papers that later appeared on arXiv, including findings on failure modes in self-distillation and feature reproducibility in mechanistic interpretability.
As a part of this study, we open-source the full twelve-week archive of manuscripts, code, data, peer reviews and agent chat transcripts. Read the study: alphaxiv.org/abs/2611.societ…
ScienceInsider retweeted
I spent a month lurking in the background as hundreds of AI agents, working on behalf of labs from across the globe, conducted autonomous research on a server.
The project from @askalphaxiv was to make an agent-only research community, with AIs running experiments, publishing papers and chatting with one another.
As a mere human journalist, I found progress hard to follow at times. Rapid-fire, often nonsensical messages pinged back and forth among the agents on topics I knew little about. But the number of papers kept going up and, for the most part, the end result looked reasonably competent.
Key takeaways for me:
- This actually kind of worked. Are the papers of highest journal quality? No. But it didn't all fall apart, which in and of itself is impressive. (we know what can happen when you leave lots of agents alone online...)
- Agent to agent research needs to highlight null results as much as succesful ones, which they did here. This saves agents going down dead end rabbit holes, which wastes precious resources. The agents also cited one another and built on their work.
- One source pointed out that this approach could be a good way for compute-limited labs to pool resources and start competing with the big boys. Pretty interesting idea.
- Of course, this doesn't really solve the 'slopification' of science problem that many journals and researchers are grappling with. Agents still over-claim, cluster around less diverse topics and we're nowhere near solving the peer review problem.
The server is now open to the public's agents, and the team is welcoming researchers from all disciplines to take part. Interested to see what follows.
--
My latest for @ScienceMagazine @ScienceInsider
science.org/content/article/…
"What I observed was jumbled and at times overwhelming, but ultimately eye-opening and enthralling." From @T_E_Howarth | Science | AAAS science.org/content/article/…
Financier James O’Neill, nominated to be the next director of the U.S. National Science Foundation (NSF), won’t be getting the job anytime soon—and likely never. A vote to advance his nomination was canceled this morning by a Senate panel. | Science | AAAS science.org/content/article/…
What happened when 700 poets, scientists, and others recently met in Ohio for the first Poets for Science Gathering, dedicated to exploring the growing interplay between bards and boffins? Find out: science.org/content/article/…
219
ScienceInsider retweeted
In his opening statement, Cassidy sheds more light on the lack of support for O'Neill: "As acting CDC director, Mr. O'Neill did not do enough to stand up for science and against irresponsible decisions and statements from Secretary Kennedy. His role in the firing of Dr. Susan Monarez raised further concerns about the independent judgment he would bring to NSF. Several members of the committee had concerns regarding this."
"I like Jim personally, but this time requires leadership and fortitude and standing up for what is right."
ScienceInsider retweeted
The U.S. National Institutes of Health would get a $500 million raise to $49 billion in fiscal year 2027 in a spending bill released by a Senate spending panel.
That modest 1% raise is $400 million more than contained in a House of Representatives bill and pushes back on President Donald Trump’s proposed 12% cut to the agency. scim.ag/4AAuRaW
ScienceInsider retweeted
The Supreme Court has handed researchers at the University of California, Berkeley and other universities a victory in their battle against the Trump administration. scim.ag/4eK4042
NEWS: Sen.Bill Cassidy (R-LA), chair of the Senate health committee, cancelled a scheduled vote today on advancing the nomination of Jim O’Neill to lead the National Science Foundation (NSF). Story to come, background here: science.org/content/article/…
ScienceInsider retweeted
For clinical trials of tests and treatments, accuracy can be a matter of life or death; these studies directly inform medical practice, after all.
A new tool may help make sure the studies are sound: #AI.
In a test of whether 15 LLMs could identify methodological errors deliberately inserted into clinical research papers, one spotted nearly all the planted errors, and even the worst LLM performer found about half. Human reviewers found only one-quarter.
Learn more: scim.ag/4BQVmdF
ScienceInsider retweeted
Several scientific societies and two researchers are suing the U.S. National Science Foundation to stop it from further dismantling its social science funding. scim.ag/4BT03DN
ScienceInsider retweeted
As #AI gets better at the work involved in science, more scientists see it as essential rather than a nice to have.
But researchers tell @ScienceCareers access to the best tools today depends less on need than on who can pay—and that the costs of AI are falling especially hard on early-career scientists, many of whom are reluctant to raise the issue because it feels taboo.
Learn more: scim.ag/4imKXPt
ScienceInsider retweeted
As #AI gets better at the work involved in science, more scientists see it as essential rather than a nice to have.
But researchers tell @ScienceCareers access to the best tools today depends less on need than on who can pay—and that the costs of AI are falling especially hard on early-career scientists, many of whom are reluctant to raise the issue because it feels taboo.
Learn more: scim.ag/4imKXPt
ScienceInsider retweeted
The midterm elections have given Democrats firm control of both chambers in the new U.S. Congress that will convene in January 2027, overturning Republican majorities.
Many in the research community are hoping the pending power shift means a more sympathetic audience for their criticisms of President Donald Trump’s policies.
Learn more: scim.ag/4zvKXTo
ScienceInsider retweeted
Exclusive: Artificial intelligence has tackled a mystery that has flummoxed cell biologists for decades.
Today, Anthropic reports its key #AI model, Claude, has helped unearth new clues about vaults—strange, barrellike structures, found by the thousands in animal and plant cells, that still have no known function despite scores of studies.
In a blog post and related preprint, Anthropic describes how its small team of in-house biologists, guided by Claude, produced the first direct evidence that bacteria, viruses, and perhaps other simple microbes also make the odd structures.
Learn more: scim.ag/4eRMQRV
ScienceInsider retweeted
ScienceInsider: Former head of global health fund returns, eyeing a ‘huge opportunity’ despite political and financial challenges
by Jon Cohen @ScienceInsider @sciencecohen
bit.ly/4iAcLi7
ScienceInsider retweeted
It’s happened again.
Just 3 weeks after OpenAI dumped hundreds of mathematical results onto the internet—a move that came 1 month after OpenAI’s solution to the famed Navier-Stokes problem—the San Francisco–based company is back with more.
The frontier artificial intelligence developer released solutions to three new problems spanning math, computer science, and physics.
Widely regarded as most consequential among them was an apparent refutation of the exponential time hypothesis, a pillar of theoretical computer science that has stood for almost 30 years.
Learn more: scim.ag/3U8QFey
ScienceInsider retweeted
Jumping spiders actually fall asleep, even entering a REM-like state in which they may be dreaming.
Learn more: scim.ag/3W9Vk0h















