Dimitra Maoutsa

The Compliance Officer Who Flagged Epstein — And Lost Her Job An FBI interview reveals Deutsche Bank’s internal warnings about Jeffrey Epstein and Jared Kushner — and alleged retaliation against the whistleblower.

”A former compliance officer for the international financial powerhouse Deutsche Bank told the FBI she was fired in 2018 after raising concerns about suspicious banking activity from accounts owned by financier and sex predator Jeffrey Epstein“ as well as Jared Kushner

A Theorist's Job is to be Wrong: A Cosmologist Explains Astrophysics is so awash in experimental data that theorists are bound to often be fruitfully wrong about what it all means.

Really nice essay by @cosmolooney.bsky.social about the importance of pursuing scientific ideas that turn out to be wrong:

"This is not a reason for despair; being wrong is not the same as failing."

🧪🔭⚛️ #cosmology
www.firstprinciples.org/article/some...

Yesterday, those who teach Intro to Sociology at Florida colleges (as opposed to universities) received a ready-made curriculum from the state and were ordered to teach it.

Yes, you read that correctly. The *state* is enforcing a curriculum on college profs, complete w/ the following restrictions:

How can RNNs learn continuously without forgetting? 🧠
Our new preprint shows how a predictive learning rule organizes recurrent dynamics into orthogonal manifolds, reducing task interference.
Congrats Zihan @zihan-liu.bsky.social !
www.biorxiv.org/content/10.6...

Together, the two papers argue:

CA1 is not simply a spatial map with task signals layered on top.

When a task demands it, temporally structured task coding can dominate -- EVEN during navigation.

In other words, space tops being the hippocampus's organizing principle.

In this revised preprint, we show that during trace learning:

- Population activity forms structured trajectories in low-dimensional space
- These trajectories reflect task structure over time
- And they align across days & animals

This builds directly on our (heavily) revised manifold paper:

Conserved hippocampal population geometry supports task generalization
www.biorxiv.org/content/10.1...

Together, they argue that task engagement can shift the dominant population structure in CA1 -- even during navigation.

New preprint out 🎉

What happens to the hippocampal “place code” when an animal is actively engaged in a task?

The answer surprised us (and might surprise you too!).

Let's dive in ⬇️

Link:
"Hippocampal trace coding dominates and disrupts place coding" www.biorxiv.org/content/10.6...

Especially science. It's so sad to be publishing so many papers that you're too busy to do science.

Large banner of Trump unfurled at Justice Department headquarters The DOJ has traditionally operated with a degree of separation from the White House to prevent the department's powers from being abused for political purposes.

Trump has turned the Department of Justice into his own personal vendetta machine. And Pam Bondi has been happy to go along for the ride.

We're witnessing the most egregious politicization of the DOJ in recent memory — and I lived through the Nixon years. www.nbcnews.com/politics/jus...

DOJ Scrubs Record of Interviews With Trump Accuser From Epstein Files The FBI interviewed one of Jeffrey Epstein’s victims four times over her allegation that Donald Trump assaulted her when she was underage.

DOJ Scrubs Record of Interviews With Trump Accuser From Epstein Files
newrepublic.com/post/206765/...

The political effects of X's feed algorithm https://doi.org/10.1038/s41586-026-10098-2 Received: 16 December 2024 Accepted: 4 January 2026 Published online: 18 February 2026 Open access • Check for updates Germain Gauthier,5, Roland Hodler?5, Philine Widmer35 & Ekaterina Zhuravskaya3,4,5 m Feed algorithms are widely suspected to influence political attitudes. However, previous evidence from switching off the algorithm on Meta platforms found no political effects'. Here we present results from a 2023 field experiment on Elon Musk's platform X shedding light on this puzzle. We assigned active US-based users randomly to either an algorithmic or a chronological feed for 7 weeks, measuring political attitudes and online behaviour. Switching from a chronological to an algorithmic feed increased engagement and shifted political opinion towards more conservative positions, particularly regarding policy priorities, perceptions of criminal investigations into Donald Trump and views on the war in Ukraine. In contrast, switching from the algorithmic to the chronological feed had no comparable effects. Neither switching the algorithm on nor switching it off significantly affected affective polarization or self-reported partisanship. To investigate the mechanism, we analysed users' feed content and behaviour. We found that the algorithm promotes conservative content and demotes posts by traditional media. Exposure to algorithmic content leads users to follow conservative political activist accounts, which they continue to follow even after switching off the algorithm, helping explain the asymmetry in effects. These results suggest that initial exposure to X's algorithm has persistent effects on users' current political attitudes and account-following behaviour, even in the absence of a detectable effect on partisanship.

A new paper shows that less than 2 months of exposure to Twitter’s algorithmic feed significantly shifts people’s political views to the right.

Moving from chronological feed to the algorithmic feed also increases engagement.

This is one of the most concerning papers I’ve read in awhile.

every leftist i know is very excited about technology but unfortunately they want tech that improves society and makes people's lives easier and that's harder for wealthy prdophiles to build a trillion dollar industry bubble around

Symmetry in language statistics shapes the geometry of model representations Although learned representations underlie neural networks' success, their fundamental properties remain poorly understood. A striking example is the emergence of simple geometric structures in LLM rep...

In our new preprint, we explain how some salient features of representational geometry in language modeling originate from a single principle - translation symmetry in the statistics of data.

arxiv.org/abs/2602.150...

With Dhruva Karkada, Daniel Korchinski, Andres Nava, & Matthieu Wyart.

She never lived to see this moment, but it’s worth remembering that none of this would have happened without Virginia Giuffre.

She had the courage to speak out against power and privilege when it was just her word against Andrew’s.

Thanks to her, justice & accountability may finally have its day.

We were told to scrub "biodiversity" from our websites because of this shit.

Memorization vs. generalization in deep learning: implicit biases, benign overfitting, and more Or: how I learned to stop worrying and love the memorization

What is the relationship between memorization and generalization in AI? Is there a fundamental tradeoff? In infinitefaculty.substack.com/p/memorizati... I’ve reviewed some of the evolving perspectives on memorization & generalization in machine learning, from classic perspectives through LLMs.

So what do we think is going on here? TL;DR: AS modes reflect a cell's engagement with various region/state-specific activity patterns. GS mode reflects E/I balanced inputs, and each neuron has a different balance point that sets its GS rate.

(see also: pubmed.ncbi.nlm.nih.gov/36086892/)

Noise is key to evolution; entropy is key to metabolism. We have to stop pretending that biology fights against these things rather than relying on them!

𝗕𝗿𝗮𝗶𝗻 𝗽𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴 𝗶𝘀 𝗿𝗲𝗺𝗮𝗿𝗸𝗮𝗯𝗹𝘆 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 𝗱𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝘁
Circuits and networks can shift "computations" quite dramatically.
Check this computational model of the role of neuromodulators.
#neuroskyence
doi.org/10.1162/Neco...

Then there was that time on study section where the grant said "If Aim 1 doesn't work we're fucked."

Added points for honesty

Deducted points for proofreading

#nihgrants

Neuronal spiking in the mammalian forebrain is dominated by a heterogeneous ground state Neuronal firing patterns have significant spatiotemporal variability with no agreed-upon theoretical framework. Using a combined experimental and mode…

With this one in print, I think I finally earned that PhD... 😅
Presented for the first time at the cosyne when the world ended (March 2020). I'll bring over a summary thread from twitter when it was still twitter...

www.sciencedirect.com/science/arti...

The revised version of our paper on the impact of top-down feedback is now out @elife.bsky.social:

doi.org/10.7554/eLif...

tl;dr: we show that using human-brain-like feedback/anatomy in a deep RNN leads to human-like visual biases!

This work was led by @tmshbr.bsky.social

#NeuroAI 🧠📈 🧪

Pavlov would have been shocked and stoked by this study!

It meticulously shows that what matters most for cue-reward learning is the duration between rewards: the longer this duration, the fewer the number of trials needed for behavioral and DA learning

www.nature.com/articles/s41...

Excited to share our new work from @kenmiller.bsky.social lab!

How do different cell types, interacting via recurrent connections, give rise to context-dependent processing and circuit stability, and what dynamical signatures reveal their individual roles? (1/11)

doi.org/10.64898/2026.02.06.704473

🧪🧠 New preprint: helping resolve a decades-long debate in synaptic plasticity

NMDA receptors are central to Hebbian learning. Yet for >30 years, the existence and function of presynaptic NMDA receptors have remained controversial.

📄 doi.org/10.64898/202...

1/6

Why motor learning involves multiple systems: an algorithmic perspective www.biorxiv.org/content/10.6... (Preprint)

Thrilled to finally share this work! 🧠🔊

Using a new reinforcement-free task we show mice (like humans) extract abstract structure from sound (unsupervised) & dCA1 is causally required by building factorised, orthogonal subspaces of abstract rules.

Led by Dammy Onih!
www.biorxiv.org/content/10.6...

Duration between rewards controls the rate of behavioral and dopaminergic learning - Nature Neuroscience Cue–reward learning rate scales proportionally with the time between rewards. Consequently, learning over a fixed duration is independent of the number of trials. This challenges trial-based dopamine ...

Very excited to post our paper led by @daburke.bsky.social www.nature.com/articles/s41... where we uncover a simple mathematical rule underlying how brains learn that a cue predicts a reward. 1/26

This is the piece I've been waiting for on how the cognitive science Epstein, Brockman and co were interested in was a particular narrow one.