New blog post: Questions about the future of AI A 6,000-word clusterfuck of considerations about economics, history, training, investment, and more. Thread of select questions below:

Had a really great chat with Scott Pelley 60 Minutes about AI & its future. I’ve always believed AI could be the ultimate tool to help us advance scientific progress and better understand the universe around us, and perhaps one day even solve all disease: youtube.com/watch?v=1XF-NG…

High school-level Math Problems with Solutions 1/21

Google announced LLMs are Greedy Agents on Hugging Face Effects of RL Fine-tuning on Decision-Making Abilities

Diffusion models generalize *really* well: if you give them a million pictures of cats, they'll learn to generate reasonable-looking cats no one's ever seen before. But the weird thing is that no one knows why they work! In a theory paper accepted to #ICLR2025, I dug into this.

You might wonder how models like OpenAI o3 are trained The key behind them is "multi-turn reinforcement learning". It means that the model is trained to perform reasoning steps, call functions, get function call results, do more reasoning, call more functions etc. all in one go

it's quite easy to calculate VRAM requirements for a large language model

5 huge AI agents, MCP, and LLM updates today: 1. China just dropped an opensource Deep Research multi-agent framework with MCP support. It connects AI Agents with specialized tools for tasks like web search, crawling, and Python code execution. 100% opensource.

MCP vs API: Simplifying AI Agent Integration with External Data youtu.be/7j1t3UZA1TY?si… 来自 YouTube

Massive release here! First, MCP. Then, A2A. Now, we have a new AI protocol. AG-UI is the Agent-User Interaction Protocol. This is a protocol for building user-facing AI agents. It's a bridge between a backend AI agent and a full-stack application. Up to this point, most

if you want to learn RL but focused on LLMs, this is the free book to read: RLHF by Nathan Lambert > RHLF training and fine-tuning > ml, nlp, rl background > preference data > reward modeling > policy gradient algorithms it’s an easy read, check it out.

Reasoning LLMs Guide Here is my practical guide to building with Reasoning LLMs. Lots of dev tips in it. It covers: - What are Reasoning LLMs? - Top Reasoning Models - Reasoning Model Design Patterns & Use Cases - Reasoning LLM Usage Tips - Limitations with Reasoning Models

12 Foundational AI Model Types ▪️ LLM ▪️ SLM ▪️ VLM ▪️ MLLM ▪️ LAM ▪️ LRM ▪️ MoE ▪️ SSM ▪️ RNN ▪️ CNN ▪️ SAM ▪️ LNN Save the list and check this out for explanations and links to the useful resources: huggingface.co/posts/Kseniase…

Though Greece may have birthed Western philosophy, Rome revitalized it. Here are the 10 greatest Roman philosophers that any true fan of Rome needs to know🧵

With Cursor/Lovable/Claude Code, the cost of software creation is approaching zero. I think this will fundamentally change software business models over the next decade. Software itself will be less differentiated, meaning differentiation will have to come elsewhere; in many

The most important AI paper of 2025 might have just dropped. NVIDIA lays out a framework for Small Language Model agents that could outcompete LLMs. Here’s the full breakdown (and why it matters):

Thanks david friedberg & the The All-In Podcast crew for inviting me to the summit, it was a blast! Fun conversation discussing everything from the limitations of today's AI systems to the latest world models like Genie 3 and their key potential role in robotics - enjoy!

Google introduces Test-Time Diffusion Deep Researcher Don't sleep on diffusion models. Test-Time Diffusion Deep Researcher (TTD-DR) is a deep research agent that models research writing as a diffusion process. Instead of static reasoning or bolted-on tools, the system drafts

Did I cross from madness to genius or vice versa? Only one way to find out, check out the latest The Cognitive Revolution Podcast with Nathan Labenz to find out!