A new study from Anthropic finds that gains in coding efficiency when relying on AI assistance did did not meet statistical significance; AI use noticeably degraded programmers’ understanding of what they were doing. Incredible.

I am very excited about this new preprint! @likeajumprope.bsky.social presents an approach for charting growth velocity using normative models. It provides a fundamental framework for detecting individualised deviations from an assumed developmental trajectory. Enjoy! arxiv.org/abs/2601.07591

Also out today in PNAS: Anwesha Das shows in this thorough paper that "micro-offline gains" in sequence learning do not reflect consolidation of a motor memory - they also happen in random sequences. With @eazanon.bsky.social and Max-Philip Stenner (Magdeburg). www.pnas.org/doi/10.1073/...

... and a lifespan normative model of fractional anisotropy from Ramona Cirstian. An important contribution! linkinghub.elsevier.com/retrieve/pii...

A bit belated, but here are some new papers! first, a conceptual framework for modelling extremes in neuroimaging data onlinelibrary.wiley.com/doi/full/10.... Great work from the inimitable Charlotte Fraza!

Come work with us! we are hiring a scientific software developer! www.radboudumc.nl/en/vacancies...

Come work with us! we are recruiting for a talented PhD candidate for a new project www.radboudumc.nl/en/vacancies...

[🆕RESEARCH] Increasing vegetation cover by 10%, 20%, and 30% could lower the global population-weighted mean temperature during the #warm season. These reductions could prevent approximately 0.86, 1.02, and 1.16 million deaths—corresponding to 27.16%, 32.22%, and 36.66% of all heat-related deaths.

... and another cool framework from Loran Knol! www.medrxiv.org/content/10.1...

Very pleased to see this paper out! An analytical frameowrk for digital phenotyping data based on hidden Markov Models. Great stuff from Imogen Leaning! jmir.org/2025/1/e64007

2/ Our first speaker was Prof. André Marquand (@amarquand.bsky.social), who studies how environmental challenges affect mental health. His talk reinforced that neuroscientists can also play a key role in researching impacts and effective ways to promote climate action engagement and policy-making! 🧠

In irony too funny for words, Anthropic asks that you please don’t use AI in your job application. No, really! open.substack.com/pub/garymarc...

A colleague today asked about Gaussian Processes on a sphere. Like spatial autocorrelation on the globe. I asked @avehtari.bsky.social. There are papers! This problem common enough to be worth having a case study I think? (1) Theory doi.org/10.1007/s112... (2) Application doi.org/10.48550/arX...

My impression of this action is that it is quite deliberately intended to sow chaos, doubt, and destruction. wapo.st/4gfgcHQ

Week 3 homework for my stats course. "Causal salad": adding measurements to a regression until something is significant and interpreting every coefficient as a total causal effect. This is rarely sensible. (All course materials here github.com/rmcelreath/s...)

New paper in Imaging Neuroscience by Hossein Rafipoor, Saad Jbabdi, et al: Hierarchical modelling of crossing fibres in the white matter direct.mit.edu/imag/article...

“LLMs amplify existing security risks and introduce new ones” Even Microsoft sees it. New paper at arXiv, discussed briefly below. open.substack.com/pub/garymarc...

🎉 eLife is pleased to announce Timothy Behrens (@behrenstimb.bsky.social) as our new Editor-in-Chief! A distinguished neuroscientist and long-time supporter, Tim will lead our efforts in transforming research communication for all.

lol, a clip of a Tesla's vision system trying to understand a train labyrinth.zone/notice/Aq5V7...

“𝗧𝗵𝗿𝗲𝗲 𝗺𝗮𝘁𝗵𝗲𝗺𝗮𝘁𝗶𝗰𝗶𝗮𝗻𝘀 𝗮𝘁 𝘁𝗵𝗲 𝗯𝗲𝗮𝗰𝗵, 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲𝗶𝗿 𝗵𝗲𝗮𝗱𝘀 𝗶𝗻 𝘁𝗵𝗲𝗶𝗿 𝘀𝗮𝗻𝗱”

Should also mention: associated OSF resource osf.io/d9r3x/ with links to code & demo videos. Again, huge tks to contributors incl. @oesteban.bsky.social @amarquand.bsky.social @engra.me @clarekelly.bsky.social @jivr.bsky.social @christiangaser.bsky.social

The projected burden of dementia in the United States www.nature.com/articles/s41...

The man in charge of social media platforms used by billions should show some moral courage, writes @jemima.bsky.social www.ft.com/content/682c...

🎉

New paper! This is a very elegant way to assess longitudinal change against normative models estimated cross-sectionally! Great stuff from Barbora Rehák Bučková! elifesciences.org/reviewed-pre...

Calibration “resolves” epistemic uncertainty by giving predictions that are indistinguishable from the true probabilities. Why is this still unsatisfying? statmodeling.stat.columbia.edu/2024/12/31/c...

o3, #AGI, the art of the demo, and what you can expect in 2025 "I saw zero evidence that o3 could work reliably in open-ended domains... The most important question wasn’t really addressed." @garymarcus.bsky.social garymarcus.substack....#AGI, #generativeAI

🛠️ November's most-read Tools and Resources paper presents a new method to identify dynamic causal interactions in complex systems: https://buff.ly/3Zldxpt Learn more about publishing your tools or resources with us: https://buff.ly/4f2O0Ht

"The media should be asking hard questions, not fanning hype." Great technical piece by Gary Marcus on OpenAI's (supposed) ARC-AGI test on the o3 model.

I know that many people are worried about eLife’s new publishing model - but please, please read this blog.

Folks who oppose elife are missing the point. Experimenting requires fiddling with the knobs and they are the only ones in town actually trying anything.

That's pretty unique behavior!

OK If we are moving to Bluesky I am rescuing my favourite ever twitter thread (Jan 2019). The renamed: Bluesky-sized history of neuroscience (biased by my interests)

Working on a case study for survival analysis. These models are odd compared to typical GLM examples: missing data (censoring) that cannot be ignored, data model (likelihood) and data-generating model not same, how to visualize predictions, 2+ equivalent ways to program. Good real data wrestling.

new preprint! The rationale and design for the PRECOGNITION study funded by the @wellcometrust.bsky.social www.medrxiv.org/content/10.1...

Since it is inefficient to attempt to educate every reviewer individually, I am yeeting into your feed this clear paper from Gelman, Hill, and Yajima on how Bayesians can do even better than correcting for multiple comparisons. arxiv.org/abs/0907.2478

Beginning to understand why so many Americans are obsessed with trans issues and panicking about being replaced. They think there are 21x more trans people, 27x more muslims, 15x more jews, 5x more asians, 3x more black people, and 2x more immigrants than there actually are.

My lab has been working on comparing neural representations for the past few years - methods like RSA, CKA, CCA, Procrustes distance We are often asked: What do these things tell us about the system's function? How do they relate to decoding? Our new paper has some answers arxiv.org/abs/2411.08197

Important new study shows that current climate models underestimate the human-caused slowing of the #AMOC (Atlantic Meridional Overturning Circulation), because they neglect freshwater influx from Greenland melt and other sources. /1 🌊 www.nature.com/articles/s41...

Want to do a PhD in Nijmegen on 'Mapping language-relevant genetic effects on brain variability using normative modelling' with myself, @amarquand.bsky.social and @profsimonfisher.bsky.social ? Applications welcome via mpi.nl/imprs-phd-fe... We are a friendly international group of🧠🧬scientists

Our atlas is now out in @NatureComms! Thank you to our reviewers whose insightful comments helped sharpen our argument. I might have had too much fun with the 33-page response - glad it lives alongside the paper. rdcu.be/dViIJ #Cerebellum #Neuroskyence #fMRI

"We might say that Twitter has reached its final form as a right-wing echo chamber, with its $44 billion purchase price essentially funding its transformation into Truth Social Premium."

I've been saying the same things for years: Why good data science cannot be automated http://blog.mikiobraun.de/2014/02/data-analysis-hard-parts.html from @mikiobraun

A machine learning approach for parsing individual variability in psychosis on the basis of cognition. A very nice contribution from Rob McCutcheon! jamanetwork.com/journals/jam...

Hey new followers! 👋 We are excitedly working away on lots of new habit research - days of training with electrophysiology 🧠 with neural decoding, causal work on making and breaking 🙅‍♀️, and 📱 assessments. As a preview - check out our latest in TiCS www.cell.com/trends/cogni...