Gary Wang @garygarywang

Introducing InfiniteDiffusion, my independent paper accepted to #SIGGRAPH2026! I have one RTX 3090 Ti. No funding, advisors, or team. By day I'm a new grad SWE at Walmart. The paper has two main contributions: - InfiniteDiffusion: a new approach to infinite generation with diffusion models. - Terrain Diffusion: the world’s first learned procedural terrain generator. Here’s why this matters, and how they are connected. 🧵

llama.cpp at 100k stars now that 90% of the code worldwide is being written by AI agents, I predict that within 3-6 months, 90% of all AI agents will be running locally with llama.cpp 😄 Jokes aside, I am going to use this small milestone as an opportunity to reflect a bit on the project and the state of AI from the perspective of local applications. There is a lot to say and discuss and yet it feels less and less important to try to make a point. Opinions about viability of local LLMs are strongly polarized, details are overlooked, the scientific approach is lacking. Arguments are predominantly based on vibes and hype waves. One thing is clear though - local LLMs are used more and more. I expect this trend to continue and likely 2026 will end up being one of the most important years for the local AI movement. I admit that I didn't expect the agentic era to come so quickly to the local LLM space. One year ago, the available models were too computationally expensive for doing long-context tasks. There wasn't an obvious path towards meaningful agentic applications. The memory and compute requirements were huge. Last summer, with the release of gpt-oss, things started to change. It was the first time we saw a glimpse of tool calling that actually works well within the resource constraints of our daily devices. Later in the year, even better models were released and by now, useful local agentic workflows are a reality. Comparing local vs hosted capabilities at a given moment of time is pointless. To try put things into perspective: - We don't need frontier intelligence to automate searches and sending emails - We don't need trillion parameter models to be able to summarize articles or technical documents - We don't need massive GPU data centers to control our home appliances or turn the lights off in the garage I believe that there is a certain level of intelligence we as humans can comprehend and meaningfully utilize to improve our working process. Beyond that level, access to more intelligence becomes unnecessary at best and counterproductive at worst. I also believe that that level of useful artificial intelligence is completely within reach locally and it has always been just a matter of implementing the right software stack to bring it to the end user. With llama.cpp, I am confident that we continue to be on the right track of building that software stack! The llama.cpp project is going stronger than ever. With more than 1500 contributors, the project keeps growing steadily. From technical point of view, I think that llama.cpp + ggml is the only solution that actually makes sense. That is, the software stack must run efficiently on every possible device, hardware and operating system. The technology is too important to be vendor-locked. It has to be developed in the open, by the community, together with the independent hardware vendors. This is the only right way to build something that will truly make a difference in the long run. I won't try to convince you about what is currently and will be possible with local AI. We will just continue to build as usual. I am confident that after the smoke clears and we look objectively at what we have built together, the benefits will be obvious to everyone. Big shoutout to all llama.cpp maintainers. I feel extremely lucky to be able to work together with so many talented contributors. Every day I learn something new and I feel there is so much more cool stuff that we are going to build. Also, I am really thankful that the project continues to have reliable partners to support it! Cheers!

keyboard_arrow_left Previous keyboard_arrow_right Next

📢 Another exciting step forward today with the launch of Gemini 3.1 Flash Live. It natively understands audio, making it much more capable of handling complex instructions. It leads on ComplexFuncBench, and on Scale AI’s AudioMultiChallenge, demonstrating skill in complex instruction following and long-horizon reasoning amidst the interruptions of real-world audio. That means the model can pick up on nuances like pitch and pace, leading to much more fluid, high-fidelity voice interactions. It’s now powering Gemini Live and Search Live globally. Huge congrats to the teams that made this possible. ✨ blog.google/innovation-and…

Say hello to Gemini 3.1 Flash Live. 🗣️ Our latest audio model delivers more natural conversations with improved function calling – making it more useful and informed. Here’s what’s new 🧵

Gemini 3.1 Flash Live is our highest quality audio & voice model yet - and a big leap towards building next-gen voice-first agents. Lower latency, better precision, more natural interactions... try it now with Gemini Live in the @GeminiApp or build with it in @GoogleAIStudio!

Google DeepMind @GoogleDeepMind

3 months ago

Say hello to Gemini 3.1 Flash Live. 🗣️ Our latest audio model delivers more natural conversations with improved function calling – making it more useful and informed. Here’s what’s new 🧵

120 214 2K 586K 557

Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI

Optimization theory for adaptive methods actually predicts most of what we know about hyperparameter scaling in LLM pretraining, and suggests new strategies as well. We did a deep dive here.

I’m very happy to present my toy research project: Sotaku! It's a neural net that automatically discovered the rules of sudoku and learned to solve them, achieving a new state-of-the-art score of 98.9% on one of the hardest sudoku datasets, while being agnostic to the game, and beating all other sudoku-optimized neural net architectures* Read more for fun motivations, plus some extremely unconventional discoveries, e.g. reverse curriculum consistently beating curriculum (!), emergent reasoning-like capabilities, and the future of traditional programming

Can language models learn useful priors without ever seeing language? We pre-pre-train transformers on neural cellular automata — fully synthetic, zero language. This improves language modeling by up to 6%, speeds up convergence by 40%, and strengthens downstream reasoning. Surprisingly, it even beats pre-pre-training on natural text! Blog: hanseungwook.github.io/blog/nca-pre-p… (1/n)

We just completed the largest decentralised LLM pre-training run in history: Covenant-72B. Permissionless, on Bittensor subnet 3. 72B parameters. ~1.1T tokens. Commodity internet. No centralized cluster. No whitelist. Anyone with GPUs could join or leave freely. 1/n

Holy moly this was my dream when Gemini started, to make video tutorials on highly technical topics. I took my old post showing Muon Shampoo relation x.com/_arohan_/statu… and gave that to NoteBookLM to make a cinematic video. It did it! Now Imagine using something like this but changing this to @karpathy giving this lecture! 🤯

rohan anil @_arohan_

4 months ago

Notebook LM cinematic video explainers. Generation has been going for an hr! Should be interesting what comes out.

2 1 17 31K 16

In huge news for global open-source AI, Qwen’s tech lead has unexpectedly stepped down, seemingly not fully of his own volition. Lin, born in 1993, was one of the driving forces behind Qwen’s early decision to go open-source with its models, even before DeepSeek forced everyone else to do the same. One of the things I enjoyed the most preparing mini-profiles of Qwen technical staff a few months back was digging into some old Lin blog posts in which he expressed so eloquently his passion for the open-source community and why he truly believed in it as an indispensable force for achieving AGI. This part for me really encapsulates the cosmopolitan makeup of the man, a PKU humanities grad who speaks six languages. “我是不喜欢把做模型这事上升高度的，也不喜欢使用国内海外的说辞。开源社区就是一个世界的社区，你的模型就应当是世界的模型。我喜欢和鼓吹开源，最重要的就是，至少我们可以站在同一个舞台上，不论输赢都能体面一些，同时，也有机会收获更多的认可。 I don’t like to elevate AI model development to some grand narrative, nor do I like the rhetoric of “Chinese vs. foreign”. The open-source community is a global community, and your model ought to be a model for the world. I love and advocate open source, and most importantly, it allows us to stand on the same stage. Win or lose, we can do it with dignity, and we also have the chance to gain greater recognition.”

Junyang Lin @JustinLin610

4 months ago

me stepping down. bye my beloved qwen.

2K 716 13K 6.6M 1K

You might have seen the WuBOT performing at the 2026 Spring Festival Gala; however, most high-dynamic extreme motions you see are executed by overfitted tracking policies. Until now, training a unified policy capable of performing various extreme motions with a high success rate remained an unsolved challenge. We spent an entire year digging into the barrier between general tracking and extreme physical behaviors. After burning through dozens of G1 robots, we finally identified the bottleneck of learning and physical executability. With these discoveries, we developed OmniXtreme: the first general policy that can execute diverse extreme motions, including consecutive flips, extreme balancing, and even breakdancing with rapid contact switches! This capability is achieved by pre-training a flow-based generative control policy and then post-training with actuation-aware residual RL for complex physical dynamics—a step we found critical for successful real-world transfer. This work is a joint collaboration with @UnitreeRobotics. Together, we are pushing the physical limits of humanoid robots. It is incredibly exciting to see a general "robot gymnast" and "robot breakdancer" come to life! It was also our first time publishing a paper with XingXing, which was an enlightening experience. The model checkpoints are now released—we welcome you to play with them! 📦 📄 Paper: arxiv.org/abs/2602.23843 🌐 Project: extreme-humanoid.github.io 💻 Code: github.com/Perkins729/Omn…

A really sad and embarrassing day to be employed at OpenAI.

Really enjoying Lyria 3 (via Gemini App). It's a fun model to create with! Glad to be at Google, and to see all the neat tracks people are making. Check out the prompt guide too if you are interested in trying it out deepmind.google/models/lyria/p…

We trained a tiny 4B model to reason for millions of tokens through IMO-level problems. Heaps excited to share our new blog post covering the full pipeline, from distilling the 🐳 to augmenting RL with a reasoning cache that unlocks extreme inference-time scaling for theorem proving. huggingface.co/spaces/lm-prov…

Used Genie 3 to recreate an early 2000s Blockbuster (the model picked the VHS itself 😲)

Free Solo: The Game Genie 3:

Running Kimi K2.5 on my desk. Runs at 24 tok/sec with 2 x 512GB M3 Ultra Mac Studios connected with Thunderbolt 5 (RDMA) using @exolabs / MLX backend. Yes, it can run clawdbot.

Thanks Benjie for visiting us! 🏅🏅🏅 generalrobots.substack.com/p/physical-int…