The modern Western lifestyle is destroying your gut. Your gut is the control center influencing your mood, immunity, weight, and even lifespan. After 15 years of researching the gut & microbiome, here's how your gut works, what’s damaging it, and how to fix it: 🧵

Okay so this is so far the most important paper in AI of the year

First week as a 3 years as a Data Scientist Data Scientist x.com/TheRoyalSerf/s…

We are excited to introduce Mercury, the first commercial-grade diffusion large language model (dLLM)! dLLMs push the frontier of intelligence and speed with parallel, coarse-to-fine text generation.

Some people today are discouraging others from learning programming on the grounds AI will automate it. This advice will be seen as some of the worst career advice ever given. I disagree with the Turing Award and Nobel prize winner who wrote, “It is far more likely that the

Microsoft releases Phi-4 reasoning models! You can now run them locally with our Dynamic GGUFs. Phi-4-reasoning-plus is only 14B parameters but performs on par with o1-mini, o3-mini and Sonnet 3.7. GGUFs: huggingface.co/collections/un…

TL;DR: We built a transformer-based payments foundation model. It works. For years, Stripe has been using machine learning models trained on discrete features (BIN, zip, payment method, etc.) to improve our products for users. And these feature-by-feature efforts have worked

I just published IQ is largely a pseudoscientific swindle (Closing the Discussion) medium.com/incerto/iq-is-…

New preprint from our lab! TrialMatchAI an AI-driven patient-to-trial matching system, with outstanding work by Majd Abdallah. It uses fine-tuned LLMs + medical chain-of-thought reasoning. #LLM #ClinicalTrials #PrecisionMedicine #TrialMatchAI #Bioinformatics arxiv.org/abs/2505.08508…

Today we are releasing ether0, our first scientific reasoning model. We trained Mistral 24B with RL on several molecular design tasks in chemistry. Remarkably, we found that LLMs can learn some scientific tasks more much data-efficiently than specialized models trained from

Stand by Me (1986) a story about growing up, growing apart, and the friends who shape us forever.

👨‍🎓🧾✨#icml2025 Paper: TabICL, A Tabular Foundation Model for In-Context Learning on Large Data With jingang, David Holzmüller and Marine Le Morvan TL;DR: a well-designed architecture and pretraining gives best tabular learner, and more scalable 1/9

Accept the situation and move on.

I'm so old I worked on that dialog :-)

|￣￣￣￣￣￣￣￣￣￣￣￣￣￣| !Push To Production On Friday| |＿＿＿＿＿＿＿＿＿＿＿＿＿＿| \ (•◡•) / \ / —— | | |_ |_

The concept that helped me go from bad models to good models: Bias and Variance. In 4 minutes, I'll share 4 years of experience in managing bias and variance in my machine learning models. Let's go. 🧵

Writing down most of it so it's ready to be turned into a potential research paper in LateX. :)

Vibe coding will be just "coding" in the near future

this is unironically the right way to build a startup

Skrub: machine-learning with dataframes Presentation at PyData London, discussing how skrub facilitates data transformations and wrangling on dataframes for machine learning and introducing the exciting "DataOps" to tune and re-apply any wrangling code