Our book on Information Theoretic Cryptography published by Cambridge University Press & Assessment is here! A confession first -- book writing is much harder than what I thought and I would not get back to it soon.
Himanshu Tyagi
2,270 posts
Researching and building general intelligence
Joined January 2015
- Loyal AI is the next trillion dollar consensus. Over the last 15 years, BFT consensus on financial ledgers enabled crazy coordination and wealth creation—BTC, ETH, SOL, DOGE, everything came from it. We now need a completely different consensus to tame AGI: the consensus of
- India is on the moon and there is an IISc equation running the controller!
- Last time I saw the Maths Olympiad culture in India up-close was more than two decades ago. I don’t think we were doing so well then, but the on-ground ecosystem was rich enough to inspire many people like me who were curious about maths. Recently, I started getting exposed toIt’s a matter of immense joy and pride that India has come 4th in its best-ever performance in the International Maths Olympiad. Our contingent has brought home 4 Golds and one Silver Medal. This feat will inspire several other youngsters and help make mathematics even more
- Introducing ROMA, the open source deep research meta agent for Sentient Chat. It is the new leader on internet search benchmark like FRAMES and SealQA, outperforming previous SOTA performance reported by Kimi and other SOTA reasoning agents with search such as perplexity andAnnouncing ROMA (Recursive Open Meta Agent): our new multi-agent framework that sets SOTA in reasoning + search. Seal-0: 45.6% FRAMES: 81.7% SimpleQA: 93.9% 🧵 Read more about how recursive coordination lets agents tackle complex queries.
- ROMA is a very simple and versatile architecture that recursively breaks complex queries into simpler ones. This method of coordinating multiple agents/tools/models is apt for deep research, long horizon tasks and boosting the power of models. This is emerging as an important[1/8] 🧵 🚀 ROMA (Recursive Open Meta Agents) v0.2.0 is here! Many exciting features have been added to streamline research/production threads: for better reliability and a builder-friendly ecosystem for high-performance recursive multi-agent systems. Stay tuned for the upcoming
- I have been dabbling with using AI for maths too. There is this COLT 2020 paper where we tried to find best way to quantize Gaussian observations for mean estimation. We couldn't establish exact optimality because we had some extra dependencies coming in our bounds. I hadWell, this time it's by Terence Tao himself: mathstodon.xyz/@tao/115306424…
- Nobody in this company knocks the doorOur newest piece of the GRID just got an upgrade 😁 GRID is the world’s largest network of intelligence, containing agents, models, data sources, frameworks, and Sentient Chat—the infrastructure that stitches it all together.
00:00 - Dobby-mini has a very unique tone and has alignment heavily modified. Yet it retains performance on instruction following and reasoning. This is a remarkable feat, more so on the upcoming 70B model!Replying to @SentientAGI3/ 𝐇𝐎𝐖 𝐃𝐎𝐄𝐒 𝐃𝐎𝐁𝐁𝐘-𝐦𝐢𝐧𝐢 𝐂𝐎𝐌𝐏𝐀𝐑𝐄 𝐓𝐎 𝐎𝐓𝐇𝐄𝐑 𝐌𝐎𝐃𝐄𝐋𝐒 𝐖𝐞 𝐚𝐫𝐞 𝐭𝐡𝐞 𝐟𝐢𝐫𝐬𝐭 𝐭𝐞𝐚𝐦 𝐭𝐨 𝐢𝐥𝐥𝐮𝐬𝐭𝐫𝐚𝐭𝐞 𝐚𝐥𝐢𝐠𝐧𝐦𝐞𝐧𝐭 𝐀𝐍𝐃 𝐫𝐞𝐭𝐞𝐧𝐭𝐢𝐨𝐧 𝐨𝐟 𝐦𝐨𝐝𝐞𝐥 𝐮𝐭𝐢𝐥𝐢𝐭𝐲. Fine-tuning models for significant changes can
- WTF?! Who did this? I thought we decided not to do it!We've arrived😈 Who's ready to make open-source AI win?
- If diffusion models drive all creative arts, we will learn that humans are not more creative than a kettle dissipating heat to boil water. A bit sad...
- This is not what I meant by dog fooding@SentientAGI needs to have a hot dog eating competition, 5 hot dogs is weak, I'll outeat @sandeepnailwal with 20 😤
- It is really really good! Will be on Sentient Chat in a month!Open-source Deep Research repo called Recursive Open Meta-Agent (ROMA), beats every existing closed-source platform (ChatGPT, Perplexity, Kimi Researcher, Gemini, etc.) on Seal-0 and FRAMES ROMA is a meta-agent framework that uses recursive hierarchical structures to solve













