Dheepa Sundaram in the video says "India's democracy is under threat and I would argue that India is a global threat..." 14/15

Cursor is almost certainly the fastest company in history to reach $500M in ARR. o3 did research and created the following graph to show how long it took each company to grow from $0 to $500M in ARR.

In case you missed it, this week we launched "DSPy: Build and Optimize Agentic Apps." In this course, you'll: ✅ Learn the fundamentals of DSPy and how to use its signature and module-based programming model to build modular, traceable, and debuggable GenAI agentic applications.

This is a great talk by Laurie Voss that gives an overview of 13 different agent protocols (including MCP and A2A) Are they meaningfully different in some way? OR are they basically the same thing? Check it out! youtube.com/watch?v=kqB_xM…

Rahul Gandhi has repeated the lies, including one based on stupid assumption that only one voting machine is there in every booth. Also, his charge of BJP winning in most seats which saw voter increase is ridiculous. Data shows that Congress won 16 seats and allies its won 37

Rahul Gandhi’s latest article is a blueprint for manufacturing fake narratives, owing to his sadness and desperation of losing election after election. Read how Rahul Gandhi lied about Maharashtra: opindia.com/2025/06/congre… Here’s how he does it, step by step: Step 1:

Vivashwan Singh Sagarika Ghose Narendra Modi Mamata Banerjee No definitive evidence confirms Mamata Banerjee directly sanctioned the Chenab Rail Bridge. The USBRL project, including the bridge, was approved in 1995 under P.V. Narasimha Rao, before her tenure as Railway Minister (1999-2000, 2009-2011). While she may have supported railway

Hindu for Human Rights❌ Hindu for Sale✅ (Thread) Tomorrow on 8th June, a seminar has been organized in Los Angeles (USA) on the topic "Hinduphobia - A conspiracy theory of Right Wingers" This event has been organized by a fraud Soros funded Hindu org- HfHR 1/23

Facts: Hindus are the most educated in USA. Also, the have they highest earning income. The Indian Community is a really good role model for the rest of society in the USA to learn by and achieve more. Because of the Indian community, the USA is advanced in many areas. 🙏

🚨BREAKING: Forecasting Principles and Practice is NOW AVAILABLE IN PYTHON! Here's everything you need to know: 🧵

Shocker! Cursor system prompts have been leaked, and it's a goldmine! The Claude system prompt incorporates several identifiable agentic AI patterns as described in "A Pattern Language For Agentic AI." Here's an analysis of the key patterns used: 1. Context Reassertion

Gemini Tip: Gemini can now understand contents from URLs. Just add the URL in your prompt, and configure the URL context tool. Very handy, also works with `/llms.txt` endpoints.

Evals are the most critical element of AI product development. But there are many misconceptions. Metrics promoted by the eval vendors, like "hallucination" or "toxicity," are ineffective and often miss domain-specific issues. No wonder 85% of AI initiatives fail (Gartner,

A study found that OpenAI’s GPT-4o appears to have been trained on copyrighted, paywalled content from O’Reilly Media books. Researchers tested GPT-4o and other models by evaluating their ability to identify verbatim text from public and private book excerpts. GPT-4o

Reinforcement learning on 100,000,000,000 observations overnight on a single TinyBox has set new SOTA on Neural MMO 3. Previous best was just under 6.0. This has an effective batch size of 3 million and a minibatch of ~180k. Star PufferLib to support and come dev with us!

BREAKING: Apple just proved AI "reasoning" models like Claude, DeepSeek-R1, and o3-mini don't actually reason at all. They just memorize patterns really well. Here's what Apple discovered: (hint: we're not as close to AGI as the hype suggests)

Pretty sick - a drag and drop ui layer on top of langgraph 👀 Build langgraph agents and workflows in a no code manner

Brilliant Paper. We need to evaluate reasoning steps separately for knowledge correctness and reasoning quality LLMs give a single right answer but hide wrong facts or sloppy reasoning. This paper scores each reasoning step, exposing which sentences supply real knowledge

Common meta-learning approaches: ▪️ Optimization-based / gradient-based ▪️ Metric-based ▪️ Model-based Meta-learning trains a model to quickly learn new tasks – even from just a few examples. Here's how these 3 approaches work: 1. Optimization-based one involves finding model

Karpathy today said Cursor for Slides needs to exist.. but it already does. When asked to "Create a detailed data-driven slide deck based on Google's recent financial filings", It created a stunning 6 page deck with graphs, diagrams IN Google theme! It's called Genspark.