Arnav Das

Ok since pinning has finally been added to this place, This is a reference post of all my excursions I try around..

"The research published today in Science Advances has found that the state of the Pacific Ocean at the beginning of a forecast plays a critical, and often overlooked, role in determining whether long-range predictions of ENSO will be accurate." www.eurekalert.org/news-release...

A 28 year old man injected himself with mercury. "The patient denied any psychiatric illness and did not disclose the motive for mercury injection." pubmed.ncbi.nlm.nih.gov/42910243/ Surprisingly he didn't get mercury poisoning (yet).

Network topology shapes consensus and vulnerability in LLM populations arxiv.org/abs/2610.31165

Now's a good time to repost this. Please share and engage. #TCSSky cstheory.meta.stackexchange.com/questions/31...

Also, I don't know if there is any way to be clearer, but: If you are a theoretical computer scientist or mathematician currently working for OpenAI, you are not one of the good guys.

OpenAI has claimed a refutation of the Exponential Time Hypothetis (ETH). Russell Impagliazzo, Mohan Paturi and Ryan Williams (@rrwilliams.bsky.social) have been asked to organize a community-led effort to evaluate this claim. More here: docs.google.com/document/d/1... h/t @gautamkamath.com

The exponential time hypothesis is apparently false: openai.com/index/buildi...

Interesting looking paper on what you can and can't identify from staggered adoption panel data. Also, I move that all important equations get subtitles:

Smart isn't the same as aware. And your brain's software is also its hardware. Three reasons why AI is not conscious and perhaps never will be www.nature.com/articles/d41... #neuroscience

Must read

We should not call others lazy—not only because it’s rude, but also because it’s almost impossible to know if someone is actually lazy or not, writes @philosophyminis.bsky.social. What is laziness anyway? A Tibetan Buddhist master explains the three kinds: buff.ly/WXYCh6C

"We shall not cease from exploration, And the end of all our exploring Will be to arrive where we started And know the place for the first time." — T.S. Eliot

Nobody has a photographic memory cognitivewonderland.substack.com/p/nobody-has...

Ask a group of people a philosophical question and it’s often the case that they'll give conflicting answers to it. We quite naturally interpret this as different people having different intuitions. But that's not the only available explanation...

A nonconvex problem can sometimes have a convex dual. Toland duality explains how, and also yields the DC algorithm. A simple example shows why this can be useful. https://www.gpeyre.com/blog/2026/10/26/dc-programming-toland-duality/

The world needs a high-quality, truly open language model. I wrote about why it’s necessary and why it’s easier to build than you might think.

We had some fairly big updates to tidymodels.org this year. I'm gonna post about them this week! The first thing you might have noticed already is that we now have a dark mode on the website! Every page and element should now work both in light and dark mode! #tidymodels

nonzerosum.games/goodhartslaw

Hm. Honestly a bit bewildered by the choice. "And I know how to read a research paper." Reading a scientific paper isn't the expertise required to build real trust by translating, visualizing, critically assessing what's actually happening in practice for laypeople www.statnews.com/2026/09/21/t...

AI models tend "to resolve cross-linguistic disagreement by treating the English judgment as authoritative, rather than acknowledging the cross-linguistic inconsistency of human language." What's the problem with that, and what should be done about it?

Working‑class academics are still struggling to fit in at university theconversation.com/working-clas...

"There's no Royal Road to geometry." —Euclid to Ptolemy (according to Proculs) "O king, for traveling through the country there are private roads and royal roads, but in geometry there is one road for all."—Menaechmus to Alexander the Great (according to Stobaeus) I've been thinking about this

these pointless chatbots are what journals are literally spending our tax money on because these publishers are leeches off the public purse and then fence off academic outputs into weird systems also Maxwell / Epstein money

Inclusion is often conditional with "a hidden cost of fitting in" and facing "an unspoken expectation to adapt: to align with dominant norms of speech, behaviour and self-presentation" + creating "distance between who people are and how they feel able to present themselves" that erodes confidence.

People with smaller social networks are more likely to use chatbots for companionship This is associated with lower well-being, especially when interactions are intensive & self disclosuring This is based on an analysis of 464,687 messages from 237 users www.nature.com/articles/s41...

If real love is caused by who the beloved is, *and* should persist, what—if any—changes in the beloved might relax the latter expectation? Cool research by Mikalonyte et al finds a hierarchy: morals > memory > physical appearance … buff.ly/YSrNOlu 1/2

Okay, let's see how this goes

Politics is largely about identity, and that is something we tend to prioritize over truth. Small wonder our reasoning is impaired if we let politics guide how we think: buff.ly/07uo086

You cannot see NAZA? Read this book instead. It is heart-wrenching. It will tell you everything there is to know about Israeli barbarism is Gaza.

We’re very excited to be organizing a 1-day workshop on “Recent advances in Regularity Lemmas (with and without calibration)” at Stanford on Tuesday, November 17! 
Program and details: regularitylemmas.github.io

God, this is scary. They are killing science in universities and are pretty open about that. What’s unclear is whether they honestly think this is good for science, or consider it to be acceptable collateral damage of hurting universities, or don’t care as long as it funnels money to corporations.

Getting into the weeds of how “all polls are models” and how statistical prediction blurs with statistical sampling when surveying society.

I'm 60+, finally, they don't think at all about me!

For decades we hear women have smaller brains therefore we're stupid (yes, look up the great term for this "neurosexism") — but now it's "crows have a small brain therefore they are smart"? We're winning and have a murder on our side, I guess. Cool.

Crows will pwn your ass on PS5.

On the same theme, but also on my radar to understand. On the near-optimality of betting confidence sets for bounded means arxiv.org/abs/2310.01547 The e-posterior royalsocietypublishing.org/rsta/article... Bandit Algorithms for Prophet Inequality and Pandora's Box arxiv.org/abs/2211.085...

If you thought Ben Lamm was a shithead, you underestimated him. He's worse. Genetically engineering any creatures, but dogs especially, to be weapons of war — it's unspeakable. www.forbes.com/sites/amyfel...

"It is good that we get to hear lived experience voices. But hearing those voices is not enough. We should listen to them without prejudice & create safe and inclusive environments where they make a difference to mental health research & mental healthcare provision." lisatpg.substack.com... #philsky

"Grief is an affliction physical and material. It roils the gut and stops the sleep. It can put a tremble in the limbs, a quiver in the voice, a restless agitation run all through. It makes the knees want to let go, the body just to sink...." dailynous.com/2026/10/22/d...

We are supposed to believe that people like musk, thiel, ackman and others are really smart but the more you listen to them, the clearer it gets that most are just uncultured idiots.

@saklein.bsky.social 💞

The problem for cybersecurity enjoyers is that there hasn’t been enough incidents to achieve the regularity and excitement of say, watching sports. Finally, with AI we have solved this

the prime minister of japan is posting about PHP

📚 Now reading: Sarah Klein (2026). Modeling as Theater and Other Questionable Research Practices Compiling some snippets here. Check out the full paper 📝 muse.jhu.edu/article/1001... Will also add this thread to my page here: irisvanrooijcogsci.com/2026/02/15/%... (h/t @olivia.science ) 1/🧵

2026 anyone. But her emails.

Classic research by Tversky and Shafir suggests it is more likely that a choice will be delayed or that new options will be sought, the more conflict there is between the present options: buff.ly/vVhs3Zb

If we're being honest, most people, especially in academia, who are dismissive of AI code tools need to hear this