Wayne Qiu
3.6K posts
Today at #NeurIPS2025, we present Titans, a new architecture that combines the speed of RNNs with the performance of Transformers. It uses deep neural memory to learn in real-time, effectively scaling to contexts larger than 2 million tokens. More at: goo.gle/3Kd5ojF
你们最不喜欢中国的什么？🤔 What do you guys dislike about China? 🤔
Generalized to a recursive DSPy program: Takes *arbitrarily long* text. Builds a ToC for it, assigns chunks to sections, and, uh, just recursively handles each section in parallel. Not pseudocode. This is really a complete general-purpose summarizer for arbitrarily long text.
Through most of history, without the existence of countries with borders, collections of people with common interests (i.e., tribes) fought to seize wealth from other tribes or defend their own. As those who won got richer and more civilized, they typically got more decadent and weaker and were eventually taken down by stronger barbarians, who were in turn brought down by subsequent generations learning to be stronger. For example, that is the story of the rise of the Roman Empire and its defeat by the Gauls as well as the rises and falls of most dynasties, and with them, the rises and declines of leadership approaches. These alternating ages of barbarianism and civility contributed to periods of war that took down the more advanced civilizations when the barbarians were strong and civilizations were weak. For a picture of where we are headed and what we should do, you can pre-order my new book, How Countries Go Broke: The Big Cycle, here: amazon.com/Principles-Inv… #howcountriesgobroke #principles
This is French reporters Justine Jankowski & Marine Zambrano. Yesterday, they both blocked me. But why? They recently did an "undercover investigation" in China that claimed Chinese factories use child & forced labor. The only problem: they are lying. And I have proof. 🧵⬇️
