amarrnaik

#AI #LLM #HuggingFace #Robotics #Engineering #TechLeadership

The "Agentic Stack" of 2026 isn't just about bigger LLMs. It’s about: 1️⃣ Open Reporting Schemas 2️⃣ Session-Level Observability 3️⃣ World Model Simulations Watch the full deep dive here: 🔗 youtube.com/watch?v=UxMZfb… What’s your biggest "Agent Fail" in production? Let's discuss ! 👇

When u move from chatbot 2 Robotic Arm or Edge AI, cost of failure is absolute In Physical AI, we need Hard Verifiers code based guardrails that prevent agents from taking high-risk physical actions, regardless of what LLM says The future: Environment-Agnostic Generalist Agent

How do we solve "Long-Horizon Tasks" (tasks that take days, not seconds)? World Models. Before an agent acts, it must "simulate" the outcome in a digital twin. If the simulation predicts a crash, the agent pivots. This "look-ahead" capability is holy grail of agentic reliability

A "Pass/Fail" score is useless for an enterprise agent. We are moving toward Session-Level Reporting. You shouldn't just see the result—you need to see the "Chain of Thought," the tool calls, the cost, and the exact step where it hallucinated. Traceability is the new Accuracy.

Princeton’s Arvind Narayanan posed killer Q : If agents r so smart, why aren't they replacing human workflows yet? gap isn't IQ; it's Severity.human makes typo. An unreliable agent deletes production db We need to measure: Consistency, Robustness, Predictability & Severity

Abhijit Ghosh (Hugging Face) dropped truth bomb: Eval reporting is less transparent Companies r "benchmark maxing"—hiding failures in fine print & cherry-picking model versions The fix? Every Eval Ever. A unified, open schema for 3rd party audits. No more "vibe-based" reporting

AI agents are "crushing" every benchmark we throw at them, yet we’ve seen zero impact on global GDP. 📉 Why? Because we are measuring Capability when we should be measuring Reliability. I just finished @HuggingFace Agentic Evals Workshop. Here is blueprint for next era of AI:

@Portfolio_Bull @grok what ranking do indian cities havve

Cities with most Millionaires 🔥 1) New York City~ 3,84,500 2) The Bay Area~ 3,42,400 3) Tokyo~ 2,92,300 4) Singapore~ 2,42,400 5) Los Angelas~ 2,20,600 6) London~ 2,15,700 7) Paris~ 1,60,100 8) Hong Kong~ 1,54,900 9) Sydney~ 1,52,900 10) Chicago~ 1,27,100 11) Milan~ 1,15,000 12) Beijing~ 1,14,300 13) Osaka~ 1,12,200 14) Shanghai~ 1,10,500 15) Toronto~ 1,08,400

Where does India stand in the global AI canvas? * 2nd in Talent. * 3rd in Operating Environment. * 68th in Infrastructure! @bsindia

Projected 2026 capital spending on AI and Data Centres, by Meta, Alphabet, Microsoft and Amazon is close to 2.1% of the US GDP. In comparison, the Apollo moonshot space program was just... 0.2%! @WSJ

Every night at 9:30 PM, a train quietly leaves Bathinda in Punjab. Its destination? Bikaner, Rajasthan. 𝗕𝘂𝘁 𝘁𝗵𝗶𝘀 𝗶𝘀 𝗻𝗼 𝗼𝗿𝗱𝗶𝗻𝗮𝗿𝘆 𝘁𝗿𝗮𝗶𝗻. It is grimly known as the “Cancer Train.” Because nearly 𝟳𝟬% 𝗼𝗳 𝗶𝘁𝘀 𝗽𝗮𝘀𝘀𝗲𝗻𝗴𝗲𝗿𝘀 𝗮𝗿𝗲 𝗰𝗮𝗻𝗰𝗲𝗿 𝗽𝗮𝘁𝗶𝗲𝗻𝘁𝘀. Every night, these patients 𝘁𝗿𝗮𝘃𝗲𝗹 𝗼𝘃𝗲𝗿 𝟯𝟮𝟱 𝗸𝗶𝗹𝗼𝗺𝗲𝘁𝗲𝗿𝘀 to reach the Acharya Tulsi Regional Cancer Institute in Bikaner. One of the few institutions in India that offers 𝗳𝗿𝗲𝗲 𝘁𝗿𝗲𝗮𝘁𝗺𝗲𝗻𝘁, 𝗳𝗼𝗼𝗱, 𝗮𝗻𝗱 𝗹𝗼𝗱𝗴𝗶𝗻𝗴 to patients and their attendants. Cancer cases are rising across India. But one state that has witnessed a particularly 𝗮𝗹𝗮𝗿𝗺𝗶𝗻𝗴 𝘀𝘂𝗿𝗴𝗲 for years is 𝗣𝘂𝗻𝗷𝗮𝗯. Experts trace the 𝗿𝗼𝗼𝘁𝘀 𝗼𝗳 𝘁𝗵𝗶𝘀 𝗰𝗿𝗶𝘀𝗶𝘀 back to the 𝗚𝗿𝗲𝗲𝗻 𝗥𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻. A movement that once turned Punjab into India’s “food bowl” and helped the country 𝗼𝘃𝗲𝗿𝗰𝗼𝗺𝗲 𝗳𝗮𝗺𝗶𝗻𝗲 𝗮𝗻𝗱 𝗵𝘂𝗻𝗴𝗲𝗿. But that success came at a cost: unchecked use of 𝗰𝗵𝗲𝗺𝗶𝗰𝗮𝗹 𝗳𝗲𝗿𝘁𝗶𝗹𝗶𝘇𝗲𝗿𝘀 𝗮𝗻𝗱 𝗽𝗲𝘀𝘁𝗶𝗰𝗶𝗱𝗲𝘀. The widespread use of these chemicals 𝗽𝗼𝗶𝘀𝗼𝗻𝗲𝗱 𝘁𝗵𝗲 𝘀𝗼𝗶𝗹, 𝗰𝗼𝗻𝘁𝗮𝗺𝗶𝗻𝗮𝘁𝗲𝗱 𝘁𝗵𝗲 𝘄𝗮𝘁𝗲𝗿, and ultimately harmed the very farmers who fed the nation. Yet, this tragedy rarely makes it to primetime debates. 𝗦𝗵𝗼𝘂𝗹𝗱 𝘁𝗵𝗮𝘁 𝘀𝘂𝗿𝗽𝗿𝗶𝘀𝗲 𝘂𝘀? Perhaps not. 𝗕𝘂𝘁 𝘀𝗵𝗼𝘂𝗹𝗱 𝗶𝘁 𝗰𝗼𝗻𝗰𝗲𝗿𝗻 𝘂𝘀? Absolutely. Because this isn’t just Punjab’s crisis. These hazardous chemicals, many 𝗯𝗮𝗻𝗻𝗲𝗱 𝗼𝗿 𝘀𝘁𝗿𝗶𝗰𝘁𝗹𝘆 𝗿𝗲𝗴𝘂𝗹𝗮𝘁𝗲𝗱 𝗶𝗻 𝘁𝗵𝗲 𝗪𝗲𝘀𝘁, are still widely used across Indian states. And the common citizen remains completely unaware. ------------------------------- This is why Acharya Prashant keeps reminding us: "The c𝗼𝗺𝗺𝗼𝗻 𝗺𝗮𝗻 𝗺𝘂𝘀𝘁 𝗳𝗶𝗿𝘀𝘁 𝗯𝗲 𝗮𝘄𝗮𝗸𝗲𝗻𝗲𝗱 to such realities. Until that happens, we will continue to be poisoned, silently, in the name of '𝗽𝗿𝗼𝗴𝗿𝗲𝘀𝘀' 𝗮𝗻𝗱 '𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁'." Because in a democracy, power ultimately lies with the people. If the people begin to question, 𝗶𝗳 𝗰𝗹𝗲𝗮𝗻 𝗳𝗼𝗼𝗱, 𝗮𝗶𝗿, 𝗮𝗻𝗱 𝘄𝗮𝘁𝗲𝗿 𝗯𝗲𝗰𝗼𝗺𝗲 𝗲𝗹𝗲𝗰𝘁𝗼𝗿𝗮𝗹 𝗶𝘀𝘀𝘂𝗲𝘀, then the leaders will have no choice but to care.

The goal isn’t bigger chatbots; it’s Objective-Driven AI that is controllable and safe by design. Is the era of LLM-dominance reaching its limit? DM me for the full deck

LLMs r trained on text—thousands of years of reading. But human learn from sensory input Hierarchical JEPA aims 2 learn "World Models" (like intuitive physics) from video & observation, not just words. This is how we get 2 reasoning & planning, not just "hallucinating" next word

Yann LeCun just dropped a bombshell at the World Model Workshop. His message to the AI community is clear: “If you are interested in Human-Level AI, don’t work on LLMs.” Scaling next-token prediction isn’t the path to true intelligence. Here’s why. 🧵

Live events are usually associated with sports, and entertainment, but *there's a fast growing market for spiritual and religious events too! #lifestyle

All you wanted to know about the India-US trade deal... in one Infographic. #IndiaUSTradeDeal #IndiaUS

Most of us grew up using just ONE type of oil for cooking Now, some suggest we should have a variety of them, each for a different purpose!