Yesterday was HUGE: I saw my baby for the first time -exactly one year after a testicular tumor was cut out of my body. MVP release is scheduled for mid July, and I strongly prefer my new Cancer ♋ to the old one 🎗️! While @DeaBabajko and I were waiting for ultrasound, we got

Thanks @_akhaliq for sharing our work!🙌 Inspired by physical laws, I'm wondering: do LLMs have reasoning laws too? So we built LoRe💥, a unified framework to formalize intrinsic reasoning patterns in LRMs! 📄Paper: https://t.co/Wa8ugS7fHk 👩‍💻Code: https://t.co/iYELNdM0ew

Ugly to see some of the smartest people on Earth wasting time arguing about a pointless definition Call it general or universal who cares, what I care about is if it can solve cancer

the gap between where you are and where you want to be is called work. it's not magic. it's not luck. it's not talent. it's work. boring, repetitive, uncomfortable work done consistently over time.

Aren't we all doomed? Just curious, what in my question indicated "Height"? @Google Mind you, this is not even AI Mode. So....

The amount of different slices the government takes is insane

opus 4.5 can in context learn and intuit the basic structure of hierarchical FSQ autoencoders trained on images. i.e it tried doing a diagonal for the 3rd here (based on examples i showed it) and it was able to do it

@rronak_ I guess I'd ask then where is the real line we want to draw, because openai could update chatgpt every week (basically is anyways) but not sure we'd call that "online" or "continuous" learning despite that potentially being the case. I'd say probably its okay - and I'd assume

not getting into a philosophical debate, but this book really changed how I see the topic and made me feel more humble. human intelligence is impressive, but calling it ‘general’ isn’t very objective. my cat would disagree. to me human intelligence is better seen as socially

Fascinated by how many great people with even greater vibes I manage to meet semi-randomly.

@beffjezos monkey see autocomplete monkey use. i am monkey

This is 100% true. “Good knees” come from good squats

Always play the long game. Nothing really matters unless it matters in the long term. The short game is a trap and no matter how big you win it will leave you feeling empty at the end.

We should start calling AI art "green art" because it's so much better for the environment than human art.

💡How to train a frontier model effectively? 1. Pretrain a gigantic MoE model from scratch using a full attention model (GQA or TPA https://t.co/5AJoEjl6oH [1]) mixed with some shortSWA ( https://t.co/tvLI2VQgWu, used in GPT-OSS) or (Higher) Linear Attention

This is the exactly the reason why Prof Wen-mei Hwu and team wrote the PMPP text book. The book teaches you core parallel programming concepts from first principles and uses CUDA as a framework to implement it.

Ana Kasparian Exposes where the money California is spending on homelessness is going It’s being funneled into NGO’s & Executives are making $500k a year each “Just experiencing what I've seen on the ground in California has made me question a lot about left wing ideology” “We

next time someone actually does solve a millennium problem they should definitely schizopost nonsense lean and LLM screenshots for a few days before dropping it that would be sooo funny

There is a certain type of autist larper that thinks being a rationalist makes them better and smarter than the rest of society and I think we should bash them with rocks

@lemire I'm a lawyer at a large law firm. I use AI in my work all the time. I can confirm that GPT-5.x Pro in particular is spectacular for legal research and analysis (but still not as good as I am). However, based on my use of the models, it is not currently possible to reliably do

I keep coming back to John Schulman’s line that “value functions are underrated.” I think we still haven’t figured out how to use them for LLM agents. For token-level LLMs, the horizon is huge. If you naively put a value function on tokens, variance blows up and credit