Dhruv Malrana (@drvdhruv) 's Twitter Profile
Dhruv Malrana

@drvdhruv

ID: 540025054

calendar_today29-03-2012 14:36:00

31 Tweet

133 Followers

170 Following

Ankur Nagpal (@ankurnagpal) 's Twitter Profile Photo

When I sold my first startup in 2020, I had no idea what to do with the money I made So I ran an A/B test: I gave half to a large private bank to invest and I invested the other half myself 5 years later, here are the results:

Brett Adcock (@adcock_brett) 's Twitter Profile Photo

A Japanese expo displayed a chained robot dog that was programmed to attack anyone who came near While the specifics of the creepy Black Mirror-like robot remain unclear, the message is evident: we need to do more on the AI safety front

Palestra (@palestrasociety) 's Twitter Profile Photo

1/ We are pleased to announce that Palestra has entered into an agreement with the Ministry of Culture of El Salvador to deploy advanced LiDAR-equipped unmanned aerial systems for the mapping and surveying of archaeological sites across the country.

1/ We are pleased to announce that Palestra has entered into an agreement with the Ministry of Culture of El Salvador to deploy advanced LiDAR-equipped unmanned aerial systems for the mapping and surveying of archaeological sites across the country.
Dhruv Malrana (@drvdhruv) 's Twitter Profile Photo

Very excited to launch Deep Cogito and the Cogito v1 models today! We started Deep Cogito last year with the belief intelligence gains will move away from pre-training into post-training, and general superintelligence will be built via breakthroughs in learning algorithms,

Aditya Agarwal (@adityaag) 's Twitter Profile Photo

1/ A year ago, I met Drishan Arora and Dhruv. They had a bold and weird idea: what if super-intelligence doesn’t require massive $$$, but instead involves rethinking fundamental assumptions? And, what if it were truly open?

1/ A year ago, I met <a href="/drishanarora/">Drishan Arora</a>  and Dhruv. They had a bold and weird idea: what if super-intelligence doesn’t require massive $$$, but instead involves rethinking fundamental assumptions?

And, what if it were truly open?
ExceptionRaised (@except_raised) 's Twitter Profile Photo

What happens when 20 curious AI researchers and builders come together. Not to compete, but to tinker, driven by respect for each other’s ideas? This weekend, @exceptionraised got to experience it first-hand 🚀 Check out all showcases below🧵

What happens when 20 curious AI researchers and builders come together. Not to compete, but to tinker, driven by respect for each other’s ideas?
This weekend, @exceptionraised got to experience it first-hand 🚀
Check out all showcases below🧵
Sanctuary AI (@thesanctuaryai) 's Twitter Profile Photo

Touch: A Game-Changer in Dexterity for Robots You can’t be responsive to the world if you can’t feel it. That’s true for people—and it's just as true for robots. We’re zeroed in on a practical, yet innovative, approach to creating general purpose robots designed to perform

Google (@google) 's Twitter Profile Photo

Say goodbye to the silent era of video generation: Introducing Veo 3 — with native audio generation. 🗣️ Quality is up from Veo 2, and now you can add dialogue between characters, sound effects and background noise. Veo 3 is available now in the Google Gemini App for Google AI Ultra

Ilir Aliu - eu/acc (@iliraliu_) 's Twitter Profile Photo

What if robots could learn real-world tasks from your perspective… without ever touching a robot? This is a system that trains robot policies using nothing but human-first, egocentric video data from smart glasses. No robots, no teleop, no sensors, just humans doing real tasks

Nathan Lambert (@natolambert) 's Twitter Profile Photo

The dominance of Chinese open models is undersold in this post. Top four open models being Chinese is one thing, but also is the full top 10 of models were pre and post train is done in house, and 18 of the top 20.

The dominance of Chinese open models is undersold in this post.

Top four open models being Chinese is one thing, but also is the full top 10 of models were pre and post train is done in house, and 18 of the top 20.
Aaron Tan (@aaronhaotan) 's Twitter Profile Photo

Introducing Lume, the robotic lamp. The first robot designed to fit naturally into your home and help with chores, starting with laundry folding. If you’re looking for help and want to avoid the privacy and safety concerns of humanoids in your home, pre-order now.

Dhruv Malrana (@drvdhruv) 's Twitter Profile Photo

We are putting out four frontier intelligence models today, each hitting heavy in its weight class. The large ones are among the smartest open models in the world right now. It has become increasingly clear that superintelligence is within reach. As we continue to iterate and

Unsloth AI (@unslothai) 's Twitter Profile Photo

You can now run the world’s most powerful Western open models locally! The hybrid reasoning 671B model matches o3 & Claude-4-Opus in performance. Trained on Llama 4 & DeepSeek-R1, Cogito-v2 has 4 variants—each setting new benchmarks. Guide + GGUFs: docs.unsloth.ai/basics/tutoria…

You can now run the world’s most powerful Western open models locally!

The hybrid reasoning 671B model matches o3 &amp; Claude-4-Opus in performance.

Trained on Llama 4 &amp; DeepSeek-R1, Cogito-v2 has 4 variants—each setting new benchmarks.

Guide + GGUFs: docs.unsloth.ai/basics/tutoria…
Deedy (@deedydas) 's Twitter Profile Photo

This is the best public resource on scaling hardware for AI, and its free. "How to Scale Your Model" is the bible from Google DeepMind that covers the math, systems and scaling laws for LLM training and inference workloads. Approachable yet thorough. Absolute Must-read.

This is the best public resource on scaling hardware for AI, and its free.

"How to Scale Your Model" is the bible from Google DeepMind that covers the math, systems and scaling laws for LLM training and inference workloads.

Approachable yet thorough. Absolute Must-read.
Drishan Arora (@drishanarora) 's Twitter Profile Photo

A small update - we had more traffic than anticipated. However, the endpoints are now scalable on Together AI for all models, including the 671B MoE. Test out the model here: together.ai/models/cogito-… (A huge thanks to the folks at Together AI for making this happen so