The team worked incredibly hard to bring the cross-platform app to life. This is just the beginning. With this core in place, we’re gearing up to launch a wave of exciting new features in the coming weeks. What an incredible first few weeks at Vercept.

Excited for everyone to try this out!! The magic of Vy is learning how it thinks. Build that “theory of mind,��� and you can stop babysitting—Vy will run for hundreds of steps and get things done!

RT if you want day-zero support for @OpenAI new open-weights model to run fast on your laptop with MLX.

Would you look at that view! #BlueGhost captured its first image on the Moon that embodies everything this bold, unstoppable Firefly team has worked so hard for over the last 3+ years. And we’re just getting started! Find out what's next for #BGM1 https://t.co/oEJhJu7KHx

Boom is supersonic. On Jan. 28, 2025 at 8:31am PST / 16:31 GMT, XB-1 broke the sound barrier. Chief Test Pilot Tristan “Geppetto” Brandenburg safely and successfully achieved supersonic speed in XB-1, the first civil supersonic jet made in America. ✅New top speed: Mach 1.122

Happening Now. Watch XB-1's supersonic test flight in real-time. Join us and see XB-1 break the sound barrier from the viewpoint of our chase aircraft.

💡Imagine a multimodal LLM that masters universal UI understanding across platforms? Here it is, 🎁 we upgrade Ferret-UI to Ferret-UI 2, a generalist model for grounded UI understanding across iPhone, Android, iPad , Webpage, and AppleTV. Check the image below for visual

What an incredible feat! 👏👏

Maybe this can be achieved with chain-of-thought, question specific prompts, in-context examples. But making a powerful reasoning model available to the world that doesn't require any bells and whistles is quite impressive and useful!

But I wasn't convinced that the reasoning behind is correct. Atleast from the response it felt like it's a fluke. So I asked to clarify, and I was pretty impressed to see the detailed reasoning.

I was excited to try the reasoning capabilities of o1-preview on this same question. It got it right!

GPT 4 fails to answer this question, and with enough prompting I might be able to get the answer correct, but without any tricks, it gets the answer wrong

My favourite question to ask all LLMs is a question that requires reasoning by elimination. Here's the question:

Winning this medal is a dream come true, not just for me but for everyone who has supported me. I am deeply grateful to the NRAI, SAI, Ministry of Youth Affairs & Sports, Coach Jaspal Rana sir, Haryana government and OGQ. I dedicate this victory to my country for their incredible

Another example where it found the paper I was looking for: One notable paper in this area is "Grounding Large Language Models in Interactive Environments with Online Reinforcement Learning" by Zheng et al. (2023) but cited a completely irrelevant link:

When describing grokking: Others describe it as "grokking" - a term borrowed from computer science that implies a sudden, intuitive understanding of a complex problem It cited https://t.co/OuMVF1p73k which doesn't mention grokking and seems unrelated to the concept of grokking

I have found @perplexity_ai citations quite misleading. Even if it answers my question correctly, the corresponding citation will often be pretty uncorrelated to the text where it was cited. Examples in 🧵

What a game! What a tournament! 👏🏆🇮🇳🐐

Our tutorial on diffusion & flows is out! We made every effort to simplify the math, while still being correct. Hope you enjoy! (Link below -- it's long but is split into 5 mostly-self-contained chapters). lots of fun working with @ArwenBradley @oh_that_hat @advani_madhu on this

Want your favourite language model to solve complex robotic tasks, or play a video game, but unsure how to switch from language to actions? Our latest preprint @Apple ablates action representations in interactive settings. 🔗: https://t.co/MoAJE5PIZ5 1 /3 🧵