Image created with GPT Image 1. Image prompt: Wearing a robe of stacked books sewn together at the spine and collar, each tome bearing the names of Black scholars and coders, a tall professor strides beneath chalkboard skylights and floating AI-generated flashcards; shot in sepia tone, light dust motes glowing — schooling the machine with ancestral wisdom.
Former Google CEO Eric Schmidt-backed FutureHouse released Finch, an AI agent for discovery in biology Currently in beta, Finch can do open-ended and directed data analysis It joins FutureHouse’s four previously announced ‘superintelligent’ AI agents https://x.com/rowancheung/status/1920018905352769783
o3 now cracks new Harvard Business School cases from the PDF, in one shot I blurred the figures to not ruin the case, but I asked the AI to figure out financials, which incorporates data scattered throughout the case. More interesting, I asked it to compare to the case’s answer. https://x.com/emollick/status/1918355078253027802
DeepSeek released Prover-V2, an open-source AI combining informal math reasoning with theorem proving With 671B params, the model solves 88.9% of problems on MiniF2F It does a ‘cold-start’ to break down proofs into subgoals before formal verification https://x.com/adcock_brett/status/1919060364655800684
We just released DeepSeek-Prover V2. – Solves nearly 90% of miniF2F problems – Significantly improves the SoTA performance on the PutnamBench – Achieves a non-trivial pass rate on AIME 24 & 25 problems in their formal version Github: https://x.com/zhs05232838/status/1917600755936018715
Me: “o3, do the first chapter of Genesis as an IKEA instruction manual and show me the flatpack,” “Fix the spelling” Me: “Now come up with an idea of what to do next” o3: “How about Genesis 2 – Garden Starter Kit?” Me: “Sure.” https://x.com/emollick/status/1918545934633357596
AI agents. Agentic AI. Agentic workflows. Agentic patterns. Agents are everywhere. But what exactly are they, and how do we build robust and effective AI applications? Excited to share “Zero to One: Learning Agentic Patterns” a guide to learn common workflow and agentic design https://x.com/_philschmid/status/1919391587315958038
3/ Stanford released an 1hr lecture on Agentic AI This 1-hour lecture will teach you everything you need to know to start building with agentic LLMs, including reflection, planning, tool use, and iterative reasoning. @Sumanth_077 https://x.com/AtomSilverman/status/1918424773010571769
Stanford released an 1 hour lecture on Agentic AI and is a must-watch for every AI enthusiast! This 1-hour lecture will teach you everything you need to know to start building with agentic LLMs, including reflection, planning, tool use, and iterative reasoning. https://x.com/Sumanth_077/status/1916494871663174056
everyone says they’re building agents. very few actually know how to do it right. @AlexReibman and I are hosting a live series — how to build, evaluate, and scale real multi-agentic systems from scratch. this week — Agents 101 (building foundations that scale) next week — Evals https://x.com/n_sri_laasya/status/1917033255620252046
UAE Rolls Out AI for Schoolkids in New Push for Sector Forefront – Bloomberg https://www.bloomberg.com/news/articles/2025-05-04/uae-rolls-out-ai-for-schoolkids-in-new-push-for-sector-forefront?embedded-checkout=true
The @huggingface LLM course has new videos! 📽️ We’ve added videos on the latest topics. Join the course and check them out! https://x.com/ben_burtenshaw/status/1919761119322804723
12/ @Damiyal216 launched his latest project: a stateful language teaching agent built with OpenAI Agent SDK, Gemini 2.0 Flash (via OpenRouter), & Chainlit UI! It personalized learning with memory & handoffs—making language learning smarter. https://x.com/AtomSilverman/status/1919066852883325320
Brett Adcock on X: “DeepSeek released Prover-V2, an open-source AI combining informal math reasoning with theorem proving With 671B params, the model solves 88.9% of problems on MiniF2F It does a ‘cold-start’ to break down proofs into subgoals before formal verification https://t.co/p8wNREkF4O https://t.co/iRoVdKaAgn” / X
https://x.com/adcock_brett/status/1919060364655800684
Most YouTube tutorials skip the most important thing about building AI agents: Iterative development. Prompt optimization, error analysis, data cleaning & evaluation are often ignored. No matter how you build AI agents, you’ll need to iterate a lot, and the above are key.” / X https://x.com/omarsar0/status/1919432255350477125
Anthropic’s AI for Science Program offers up to $20,000 in API credits for a 6-month period to researchers attached to research institutions working on high-impact scientific projects, with a particular focus on biology and life sciences applications, with selections made on the https://x.com/btibor91/status/1919428974142489042
Introducing Anthropic’s AI for Science Program \ Anthropic https://www.anthropic.com/news/ai-for-science-program
Learn to build conversational AI voice agents in “Building AI Voice Agents for Production”, created in collaboration with @livekit and @realavatarai, and taught by @dsa (Co-founder & CEO of LiveKit), @shayneparlo (Developer Advocate, LiveKit), and @nedteneva (Head of AI at https://x.com/AndrewYNg/status/1920161212312268988
New short course ➡️ Building AI Voice Agents for Production LLMs can write and reason, but getting them to talk in real time, with low latency, and in a way that actually feels human, is a different challenge. In this course, created with @LiveKitAgent and @realavatarai, you’ll https://x.com/DeepLearningAI/status/1920153317562323095
(1) Voice AI Masterclass — Kwindla Hultman Kramer and swyx – YouTube https://www.youtube.com/watch?v=AbToUiWRhn4&t=972s
Rewriting Pre-Training Data Boosts LLM Performance in Math and Code Introduces two openly licensed datasets: 1. SwallowCode (≈16.1 billion tokens) refines Python snippets from The-Stack-v2 2. SwallowMath (≈2.3 billion tokens) enhances Finemath-4+ by removing boilerplate, https://x.com/iScienceLuvr/status/1920056647822532752
A fantastic and free 200+ page book covering all fundamentals of Large Language Models. https://x.com/rohanpaul_ai/status/1919304300636815782
What cyborg work looks like for an academic.” / X https://x.com/emollick/status/1917964431432290352
Cursor is now free for students. Enjoy!” / X https://x.com/cursor_ai/status/1919846420234031146
BOOOM! Learn VLMs from inside out in < 1000 lines of pure PyTorch code! 🔥 https://x.com/reach_vb/status/1919771435775533350
I warned about the Homework Apocalypse in 2023. It happened as predicted. There is a world where AI & traditional education get along very well (mixes of active in-class learning, AI-assisted assignments & tutors, blue books), but it needs to be built. https://x.com/emollick/status/1920184969852244173
I have taught an entrepreneurship class for 15 years. I just had an online Q&A with a group of entrepreneurs taking a similar class. As an experiment, I put their questions into o3. The answers were all very good. (The examples here are not from the students, but are typical) https://x.com/emollick/status/1919391279261032864
People are likely massively lying about how much they use AI In this survey, 39% of students said they never used AI and another 24% say they only use it a little… …but if you ask them how many of their peers use AI less than two days a week, they answer only 1% of them do! https://x.com/emollick/status/1919231306321322075
Try Little Language Lessons, learning experiments using Gemini 2.0. https://blog.google/outreach-initiatives/education/little-language-lessons/
Open Letter – In the age of AI, we must prepare our children for the future — to be AI creators, not just consumers. A basic foundation in computer science and AI is crucial for helping every student thrive in a technology-driven world. Without it, they risk falling behind. https://csforall.org/unlock8/open-letter
A major mistake I made in my undergrad is that I focused way too much on mathematical lens of computing – computability, decidability, asymptotic complexity etc. And too little on physical lens – energy/heat of state change, data locality, parallelism, computer architecture. The” / X https://x.com/karpathy/status/1919647115099451892
We just completed preliminary evaluations for Gemini 2.5 Pro on FrontierMath! We used an older version of our scaffold, so this is not exactly comparable to our other results. Gemini 2.5 Pro got 13% correct (±2%), compared to o4-mini’s 16% to 19% (±2%) with the same scaffold. https://x.com/EpochAIResearch/status/1918330845112262753
ChatGPT Edu now available to every medical and graduate student at @IcahnMountSinai: https://x.com/gdb/status/1919138017475723655
DeepSeek quietly released Prover-V2, an open-source AI combining informal math reasoning with theorem proving —671B params —Solves 88.9% of problems on MiniF2F —Does ‘cold-start’ to break down complex proofs into subgoals before formal verification https://x.com/rowancheung/status/1917844254648324388
I hope we can empower everyone to build with AI. Starting from K-12, we should teach every student AI enabled coding, since this will enable them to become more productive and more empowered adults. But there is a huge shortage of computer science (CS) teachers. I recently spoke” / X https://x.com/AndrewYNg/status/1917985792607363189
Dolphin-Math Datagen quickly easily creates as many math problems as you want to train your model. It is intended to be trained both SFT and RL. This teaches the model to solve math step by step, long form, by hand, just like a human does. inspired by @FernandoNetoAi It is https://x.com/cognitivecompai/status/1919626890106589570
We just published a blog post on our experience improving math reasoning in LLMs using reinforcement learning. Check it out if you’re curious about RL. https://x.com/denisyarats/status/1919601674676588894
