A prediction about our AI future in 1964:

Introduction to Machine Learning by J. Deng and R. Fong and V. Ramaswamy Lecture notes: princeton-introml.github.io

10 AI and Machine Learning textbooks (in downloadable PDF format) 1️⃣ Understanding Machine Learning, covers theory and algorithms, top pick for beginners 🔗 cs.huji.ac.il/~shais/Underst… 2️⃣ Mathematical Foundations of Machine Learning, start here if your math skills need a boost 🔗 mml-book.github.io/book/mml-… 3️⃣ Mathematical Analysis of Machine Learning Algorithms 🔗 tongzhang-ml.org/lt-book/lt-… 4️⃣ Theoretical Principles of Deep Learning 🔗 arxiv.org/pdf/2106.10165 5️⃣ Neural Networks and Machine Learning 🔗 arxiv.org/pdf/1901.05639 6️⃣ Graph Deep Learning 🔗 yaoma24.github.io/dlg_book/d… 7️⃣ The Algorithmic Perspective on Machine Learning 🔗 people.csail.mit.edu/moitra/… 8️⃣ Probability Theory: Theory and Examples 🔗 sites.math.duke.edu/~rtd/PTE… 9️⃣ Fundamentals of Applied Probability 🔗 sites.math.duke.edu/~rtd/EP4… 🔟 Advanced Data Analysis 🔗 stat.cmu.edu/~cshalizi/ADAfa…

50 legendary internet rabbit holes you can spend hours exploring👇 1. zoom.earth — Watch the world via live satellite imagery 2. flightradar24.com — See every plane currently in the sky 3. marinetraffic.com — Track all ships at sea in real time 4. windy.com — Live map of winds and storms 5. lightningmaps.org — Watch lightning strikes hitting Earth in real time 6. earthquake.usgs.gov — Live list of recent earthquakes 7. submarinecablemap.com — Ocean cables carrying the internet 8. globalforestwatch.org — Watch forests disappear from space 9. worldometers.info — The world's statistics, second by second 10. internetlivestats.com — Current number of tweets and searches being posted 11. thetruesize.com — Compare the true sizes of countries 12. oldmapsonline.org — Maps from centuries ago 13. davidrumsey.com — Archive of 150,000 historical maps 14. openstreetmap.org — World map drawn by volunteers 15. window-swap.com — Look out the window of a random person around the world 16. virtualvacation.us — Virtual walks through cities 17. mapcrunch.com — Teleport to a random spot on Earth 18. atlasobscura.com — Catalog of the world's strangest places 19. neal.fun — Interactive knowledge experiences 20. htwins.net/scale2 — Scale journey from atom to universe (Updated HTML5 link) 21. eyes.nasa.gov — Explore the solar system in 3D 22. stellarium-web.org — Real sky map in your browser 23. apod.nasa.gov — NASA's astronomy picture of the day 24. images.nasa.gov — NASA's entire visual archive, free 25. pudding.cool — Visual articles told through data 26. ourworldindata.org — The state of the world with real data 27. gapminder.org — What we mistakenly think we know about the world 28. informationisbeautiful.net — Visualizing complex data 29. data.worldbank.org — World Bank's open data 30. data.tuik.gov.tr — Turkey's official statistics database 31. archive.org — Archive of millions of books, films, and software 32. gutenberg.org — 70,000 free books whose copyrights have expired 33. openlibrary.org — Record of every book in the world 34. loc.gov — U.S. Library of Congress digital archive 35. europeana.eu — Europe's cultural heritage archive 36. dp.la — America's digital library collection 37. artsandculture.google.com — Tour museums from home 38. rijksmuseum.nl — Download artworks in high resolution 39. wikiart.org — Archive of 250,000 artworks 40. publicdomainreview.org — Forgotten visual treasures of history 41. openculture.com — Free archive of culture and education 42. metmuseum.org — Met Museum's open collection 43. musicmap.info — Family tree of music genres 44. radiooooo.com — Pick a country and decade to listen to that era 45. listen.hatnote.com — Turn Wikipedia edits into audio 46. wikipedia.org — Random knowledge well 47. timeanddate.com — Time, sunrises, and sky events 48. sciencedaily.com — Live stream of science news 49. arxiv.org — Free preprints of scientific papers 50. observablehq.com — Visualize data with live code Save this. You’ll definitely need some of these later. 🔖 Follow @Orion_Vers7x for more useful websites, AI tools & tech resources.

We just launched the DeepMind Institute, a platform for researchers across @GoogleDeepMind and the wider research community to publish and debate how increasingly capable AI should be built, governed and used. First 4 essays on reasoning transparency, economic policy and human flourishing at institute.deepmind.com.

Today @Google, we published work led by @m_codreanu on AI in Science, drawing on 15 million Gemini interactions, 2,600 specialised AI models & a 600-scientist survey. What did we find about how AI is changing science? Read on...

"From RLHF to Direct Alignment" is a recent paper that discusses the mathematics behind different approaches to aligning LLMs with human preferences. It covers methods ranging from Reinforcement Learning from Human Feedback (RLHF) to more recent approaches such as DPO, IPO, KTO and SimPO, and also looks at problems such as reward overoptimization, length hacking, mode collapse, and likelihood displacement. I found it really interesting to see how much mathematics goes into shaping the behavior of modern LLMs. Definitely worth a look. arxiv.org/pdf/2601.06108

[Download 333-page PDF. Updated version.] Mathematical Theory of Deep Learning: arxiv.org/abs/2407.18384

OpenResearch was the #1 trending GitHub repo on Friday 🚀 With OpenResearch, you can turn any coding agent into a research agent that reviews literature, develops hypotheses, runs experiments, and produces research artifacts. Own and automate your research stack end-to-end. Now available on Windows. Check it out: github.com/alphaXiv/OpenRese…

Terence Tao: The math behind today’s LLMs is actually simple. Training and running them mostly uses linear algebra, matrix multiplication, and a bit of calculus, material an undergraduate can handle. We understand how to build and operate these models. The real mystery is why they work so well on some tasks and fail on others, and why we cannot predict that in advance. We lack good rules for forecasting performance across tasks, so progress is largely empirical. A key reason is the nature of real-world data. Pure noise is well understood, perfectly structured data is well understood, but natural text sits in between, partly structured and partly random. Mathematics for that middle regime is thin, similar to how physics struggles at meso-scales between atoms and continua. Because of this gap, we can describe the mechanisms but cannot yet explain capability jumps or give reliable task-level predictions. That mismatch, simple machinery versus hard-to-predict behavior, is the core puzzle. ---- Video from Prof @Briankeating YT Channel (Link in comment)

Anthropic pays $750,000+ a year for engineers who can build LLM architectures from scratch. Stanford taught the entire thing in 1 hour lecture & released it for free. Bookmark & watch this today before someone takes it down ...

“The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement” This paper argues that true recursive self-improvement isn’t just AI getting better at tasks, but AI getting better at getting better. They map a path from AI simply executing human-designed improvements to choosing its own improvement strategies, generating its own learning experiences, adapting from deployment, and eventually modifying the mechanisms that create future improvements. The endgame is moving from humans building each better AI to building an AI whose improvements persist, feed into the next generation, and recursively improve the improvement process itself. alphaxiv.org/abs/2609.11873

Self-improving agents are a top research topic right now. This new survey is a good map of the area. (bookmark it) It splits recursive self-improvement into stages of autonomy. An agent first executes improvements someone else designed. Then it chooses its own improvement strategy, collects its own experience, adapts to new environments, and finally improves the process of improvement itself. That staging makes claims easier to check. When a paper says its agent is self-improving, you can ask which of these stages it actually automates. The survey also uses a Headroom-Closed Index to show where current LLMs fall short, and compares requirements across scientific discovery, embodied agents and software engineering. Paper: academy.dair.ai/papers/the-l…

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so. Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training. You can read the full post here: darioamodei.com/post/we-must…

so many cool classes coming up in the fall quarter! - Deep Learning Alchemy by @kothasuhas @tatsu_hashimoto @wen_kaiyue : deep-learning-alchemy.github… - Engineering AI Agents by @Diyi_Yang @michaelryan207 @jyangballin : cs329z.stanford.edu/ - AI Agents by @gneubig @dan_fried : cmu-agents.com/ - World Models by @thoma_gu : cis.upenn.edu/~cis6280/ have fun learning my friends

With AI Engineering skills, you actively shape the build: You influence what gets built, and drive the build loop. Here're key skills to do this.

🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient. 🔹 Introducing the smallest model in our new architecture family, with native visual understanding. 🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models. 1/6

10,000 OPENAI AGENTS SOLVED A $1 MILLION PROBLEM IN 88 HOURS For nearly 90 years, mathematicians could not determine whether the Navier-Stokes equations can break down while describing fluid motion. -> Now OpenAI says it has found the answer. The company deployed around 10,000 agents powered by an internal model significantly more capable than GPT-6 Astra. They worked in parallel, exchanged 2.7 million messages and generated roughly 130 billion output tokens. The digital team produced a proof in 88 hours. The system showed that an initially motionless fluid can develop a singularity under certain conditions. Its velocity grows without bound within a finite amount of time. GPT-6 Astra then spent another 17 hours formalizing and verifying the result in Lean. The computation cost millions of dollars, but the company does not plan to claim the prize. The proof must also survive independent scrutiny from mathematicians. OpenAI just demonstrated a new model for scientific discovery. A single AI is no longer trying to replace one scientist. Thousands of agents can now form a digital research lab and compress decades of human work into days ↓

I put together a mega write-up on GPT-6 Astra & looped transformers. How looped transformers / recurrent depth works, cost-tradeoffs, whether it hides reasoning traces, with lots of figures and a tour of recent looped transformer research.

Former Anthropic employee Jacob Coxon says the AI industry is "gambling with our lives." He tells Anderson what he finds "most scary is if AI is used to make itself more intelligent" and warns these companies are "compelled to race toward building a deadly technology."