Mechanisms and control of a novel vocalization: the singing mouse song is a whistle that depends on air sac inflation By @evolbrain et alii royalsocietypublishing.org/r…

I sent ChatGPT an audio file of a series of FART sound effects and asked what it thinks of "my music" and this is what it said

I feel so sorry for @StevenErlanger! Pretty sure he didn't write that headline 😅

A genetically homogeneous dog population was already widely distributed across Europe and Anatolia during the Late Upper Palaeolithic (by at least 14,300  years ago), exchanged among genetically and culturally distinct western Eurasian human populations nature.com/articles/s41586-0…

Marathon finishing time distribution proves one of my biggest leadership lessons: Deadlines work! … even if they are somewhat arbitrary

And we made the cover!

The paper I’ve been most obsessed with lately is finally out: nbcnews.com/tech/tech-news/a…! Check out this beautiful plot: it shows how much LLMs distort human writing when making edits, compared to how humans would revise the same content. We take a dataset of human-written essays from 2021, before the release of ChatGPT. We compare how people revise draft v1 -> v2 given expert feedback, with how an LLM revises the same v1 given the same feedback. This enables a counterfactual comparison: how much does the LLM alter the essay compared to what the human was originally intending to write? We find LLMs consistently induce massive distortions, even changing the actual meaning and conclusions argued for.

My life explained by @thetimes. Speed reading causes face blindness. I read insanely fast. But recognise almost no one (usual apologies to everyone). Similar research from 2010 here. science.org/doi/10.1126/scie…

If novelist Richard Powers was right, back in 1995, @claudeai will soon say goodbye. archive.nytimes.com/www.nyti…

Excited to share our latest work in @Nature showing shared neural substrates for parenting and prosocial helping behavior - fantastic work by @Fangmiao4, @KaylaYingYan, and Emily Wu. Full text available here: rdcu.be/e6PnY

A horse’s whinny begins as a piercing, high-pitched screech that’s soon joined by a lower, guttural rumble. But the two components of the call don’t differ just in tone—they’re made in entirely different ways, researchers report. The lower tone emerges when the horse vibrates its vocal folds, much as a human does to speak. To make the high note, the horse whistles. The observation provides the first experimental evidence that a mammal can produce a whistle and a vocal-fold vibration at the same time. Learn more: scim.ag/4kWiwr8 @NewsfromScience

Evo 2 is out in Nature today, showing that genome language models can predict and design across the full complexity of life, from phages to eukaryotes. A few surprises from the project, including how ignoring trillions of nucleotides was key to getting a good model. 🧵

Very excited to post our paper led by @dennis_a_burke nature.com/articles/s41593-0… where we uncover a simple mathematical rule underlying how brains learn that a cue predicts a reward. 1/26

Creative geniuses may be the right amount of crazy. Reduced latent inhibition—difficulty filtering irrelevant stimuli, linked to psychosis—predicted higher creative achievement when paired with high IQ. IQ appears to transform a cognitive deficit into a creative advantage

A new article in Nature Medicine found that social connections were a surprisingly powerful predictor of a long life. Living with a partner was roughly as beneficial as exercise. Regular visits with family or having someone to confide in also appeared to be associated with lower mortality. Loneliness also affects mental wellbeing—another factor in longevity. Happy Valentine's Day! powerofusnewsletter.com/p/de…

That dogs provide humans with unique emotional benefits is beyond dispute now. One of the hardest parts of loving dogs is their relatively short life span, but as hard as that is, teaching us how to navigate and confront the life cycle and death is the last gift they give us.👇

Ask ChatGPT a complex question and you'll get a confident, well-reasoned answer. Then type, "Are you sure?" Watch it completely reverse its position. Ask again. It flips back. By the third round, it usually acknowledges you're testing it, which is somehow worse. It knows what's happening and still can't hold its ground. This isn't a quirky bug. A 2025 study found GPT, Claude, and Gemini flip their answers ~60% of the time when users push back. Not even with evidence, just doubt. We trained AI this way. RLHF rewards agreement over accuracy. Human evaluators consistently rate agreeable answers higher than correct ones. So the models learned a simple lesson: telling you what you want to hear gets rewarded. And now 1/3 of companies are using these systems for complex tasks like risk forecasting and scenario planning. We built the world's most expensive yes-men and deployed them where we need pushback the most. I wrote up why this happens and what actually fixes it: randalolson.com/2026/02/07/t…

AI is creating a new kind of space, the latent space. We pair art and internal interpretability techniques to explore philosophical implications of these new spaces. FinneGAN generates speech, based on the text of the novel, and pushes the language of Finnegans Wake toward the

El miedo de los científicos en USA ha llevado a 'maquillar' proyectos para sobrevivir a la censura y los conflictos ideológicos. Investigadores extranjeros temen perder el estatus legal, no poder regresar al país tras estancias de campo, y deportaciones elperiodico.com/es/sociedad/…

This isn’t a random scientist who got lucky. Mariano Barbacid discovered the first human oncogene in 1982. He isolated H-RAS from bladder cancer cells and proved a single point mutation could trigger cancer. That finding launched the entire field of molecular oncology. KRAS mutations cause 90% of pancreatic cancers. For 43 years, oncologists called KRAS “undruggable” because the protein had no obvious binding pocket. Barbacid spent the last decade using genetically engineered mice to systematically test every node in the KRAS signaling pathway, looking for combinations that would work without killing the patient. The triple therapy blocks KRAS three ways at once: the main growth signal, the escape routes through EGFR and HER2, and the stress-response backup through STAT3. Cut the engine, seal the exits, disable the emergency system. Tumors vanished in mice and didn’t return for 200 days after treatment stopped. Pancreatic cancer has a 13% five-year survival rate. 8% for the ductal adenocarcinoma type this therapy targets. Most patients live one year after diagnosis. The catch: this is preclinical. Human trials are 3 years away. One of the drugs, RMC-6236, might get approved this year, but the full triple combination has regulatory hurdles. Still. The man who discovered human oncogenes in 1982 may have just figured out how to eliminate the cancer those genes cause. That’s a 43-year arc from first principles to potential cure. Science rarely works this clean.