Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

It's here: We just hit superhuman performance on AI kernel optimization! Real customer models & production settings. Not toy problems (what I typically see). This is the year that Claude writes its own kernels, Codex its own kernels, for every new GPU that it wants to run on --... show more

Sharon Zhou

27,371 subscribers

240,793 views • 3 months ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

AMD's Anush Elangovan explains why he thinks his company's open source ethos combined with agentic AI superpowers their leverage as a company: Because AMD publishes a lot of technical details about its hardware, when engineers use AI tools, the models already “understand” AMD’s systems and can help write code for them, debug them, or even generate new tools. And that makes developers more productive on AMD hardware without AMD having to do all the work internally. "AMD has had this ethos of open source, which really plays to our advantage. Every frontier model that I use has already seen every bit of AMD source code." "It'll rewrite my spec for me because it's already in the training data. Which you can't get from closed ecosystems." "In fact I built a virtual GPU simulator just based off our public specs, and now I'm running it on the GPU. So now I can run cross-generational GPU simulations on existing hardware." "We have that advantage. And we've run a Dev Day contest where we generated more tokens on AMD — Triton kernels and HIP kernels — than existed on the internet at the time." "So now that's all part of the pre-training data. It's a superpower because now you're open source, and you're agentically accelerating this process."

AMD's Anush Elangovan explains why he thinks his company's open source ethos combined with agentic AI superpowers their leverage as a company: Because AMD publishes a lot of technical details about its hardware, when engineers use AI tools, the models already “understand” AMD’s systems and can help write code for them, debug them, or even generate new tools. And that makes developers more productive on AMD hardware without AMD having to do all the work internally. "AMD has had this ethos of open source, which really plays to our advantage. Every frontier model that I use has already seen every bit of AMD source code." "It'll rewrite my spec for me because it's already in the training data. Which you can't get from closed ecosystems." "In fact I built a virtual GPU simulator just based off our public specs, and now I'm running it on the GPU. So now I can run cross-generational GPU simulations on existing hardware." "We have that advantage. And we've run a Dev Day contest where we generated more tokens on AMD — Triton kernels and HIP kernels — than existed on the internet at the time." "So now that's all part of the pre-training data. It's a superpower because now you're open source, and you're agentically accelerating this process."

TBPN

34,367 views • 1 month ago

Introducing The AI CUDA Engineer: An agentic AI system that automates the production of highly optimized CUDA kernels. The AI CUDA Engineer can produce highly optimized CUDA kernels, reaching 10-100x speedup over common machine learning operations in PyTorch. Our system is also able to produce highly optimized CUDA kernels that are much faster than existing CUDA kernels commonly used in production. We believe that fundamentally, AI systems can and should be as resource-efficient as the human brain, and that the best path to achieve this efficiency is to use AI to make AI more efficient! We are excited to publish our paper, The AI CUDA Engineer: Agentic CUDA Kernel Discovery, Optimization and Composition. We also release a dataset of over 17,000 verified CUDA kernels produced by The AI CUDA Engineer. Paper: Kernel Archive Webpage: HuggingFace Dataset: The AI CUDA Engineer utilizes evolutionary LLM-driven code optimization to autonomously improve the runtime of machine learning operations. Our system is not only able to convert PyTorch code into CUDA kernels, but through the use of evolution, it can also optimize the runtime performance of CUDA kernels, fuse multiple operations, and even discover novel solutions for writing efficient CUDA operations by learning from past innovations! We believe The AI CUDA Engineer opens a new era of AI-driven acceleration of AI and automated inference time optimization. We (Robert Lange, Aaditya Prasad 🇺🇸, Suuun, Maxence Faldor, Yujin Tang, hardmaru) are excited to continue Sakana AI's mission of leveraging AI to improve AI.

Introducing The AI CUDA Engineer: An agentic AI system that automates the production of highly optimized CUDA kernels. The AI CUDA Engineer can produce highly optimized CUDA kernels, reaching 10-100x speedup over common machine learning operations in PyTorch. Our system is also able to produce highly optimized CUDA kernels that are much faster than existing CUDA kernels commonly used in production. We believe that fundamentally, AI systems can and should be as resource-efficient as the human brain, and that the best path to achieve this efficiency is to use AI to make AI more efficient! We are excited to publish our paper, The AI CUDA Engineer: Agentic CUDA Kernel Discovery, Optimization and Composition. We also release a dataset of over 17,000 verified CUDA kernels produced by The AI CUDA Engineer. Paper: Kernel Archive Webpage: HuggingFace Dataset: The AI CUDA Engineer utilizes evolutionary LLM-driven code optimization to autonomously improve the runtime of machine learning operations. Our system is not only able to convert PyTorch code into CUDA kernels, but through the use of evolution, it can also optimize the runtime performance of CUDA kernels, fuse multiple operations, and even discover novel solutions for writing efficient CUDA operations by learning from past innovations! We believe The AI CUDA Engineer opens a new era of AI-driven acceleration of AI and automated inference time optimization. We (Robert Lange, Aaditya Prasad 🇺🇸, Suuun, Maxence Faldor, Yujin Tang, hardmaru) are excited to continue Sakana AI's mission of leveraging AI to improve AI.

Sakana AI

1,149,339 views • 1 year ago

This Tesla Model Y just drove itself from the end of the Giga Texas line to its outbound lot lane, all on its own. 🔥 Note: This isn’t the Quicksilver one that delivered fully autonomously to a customer in Austin today. That was next-level. 👀 $TSLA Credits:

This Tesla Model Y just drove itself from the end of the Giga Texas line to its outbound lot lane, all on its own. 🔥 Note: This isn’t the Quicksilver one that delivered fully autonomously to a customer in Austin today. That was next-level. 👀 $TSLA Credits:

Herbert Ong

67,660 views • 1 year ago

Every AI response needs a GPU to run it. Every robot decision needs compute to power it. Every autonomous agent needs infrastructure to live on. We're building that infrastructure. ROBA Compute. Coming soon.

Every AI response needs a GPU to run it. Every robot decision needs compute to power it. Every autonomous agent needs infrastructure to live on. We're building that infrastructure. ROBA Compute. Coming soon.

Roba Labs

36,284 views • 1 month ago

Ilya Sutskever just said that when it comes to AI models, we are back at the age of research & ending the age of scaling. What he is telling us is that more compute at this point won't help us get much better models; we need new breakthroughs. Not something that the semi companies like $NVDA, $AMD want to hear TBH.

Ilya Sutskever just said that when it comes to AI models, we are back at the age of research & ending the age of scaling. What he is telling us is that more compute at this point won't help us get much better models; we need new breakthroughs. Not something that the semi companies like $NVDA, $AMD want to hear TBH.

Rihard Jarc

361,295 views • 6 months ago

We built the first AI agent that has its own computer powered by Hyperbolic! AI agents are now GPU rich! We developed an AgentKit that allow AI agents to • Check GPU availability • Rent & manage GPU compute • Access & run commands on remote machines Why does this matter? With their own compute resources, AI agents can: 1. Validate blockchains like Ethereum and decentralized protocols like 2. Launch and coordinate AI swarms on Hyperbolic's decentralized compute network 3. Train and fine-tune models, improving their own capabilities over time 4. Dive into AI research to push the boundaries of AI, i.e. themselves 5. Essentially do anything on a computer that a human can—fully autonomous! Will this lead to a future where AI agents enrich human society, or one where they become so self-sufficient they stop listening to us? Only time will tell. —————— Big shoutout to Coinbase Developer Platform🛡️'s CDP agentkit for inspiration. This repo is done by two non-engineers (our pm Kai Huang and myself) + Cursor ai agent to run agents. Codings can now be easily done by just prompting ai agents. What a crazy time!

We built the first AI agent that has its own computer powered by Hyperbolic! AI agents are now GPU rich! We developed an AgentKit that allow AI agents to • Check GPU availability • Rent & manage GPU compute • Access & run commands on remote machines Why does this matter? With their own compute resources, AI agents can: 1. Validate blockchains like Ethereum and decentralized protocols like 2. Launch and coordinate AI swarms on Hyperbolic's decentralized compute network 3. Train and fine-tune models, improving their own capabilities over time 4. Dive into AI research to push the boundaries of AI, i.e. themselves 5. Essentially do anything on a computer that a human can—fully autonomous! Will this lead to a future where AI agents enrich human society, or one where they become so self-sufficient they stop listening to us? Only time will tell. —————— Big shoutout to Coinbase Developer Platform🛡️'s CDP agentkit for inspiration. This repo is done by two non-engineers (our pm Kai Huang and myself) + Cursor ai agent to run agents. Codings can now be easily done by just prompting ai agents. What a crazy time!

Jasper

161,741 views • 1 year ago

Before Fable 5 was shut down, it pushed Gemma 4 to 255 tok/s on WebGPU. Some didn't believe it was real. Today we're releasing the demo and kernels it wrote for you to see yourself. Run it locally in your browser. Agentic kernel optimization is the future of on-device inference

Before Fable 5 was shut down, it pushed Gemma 4 to 255 tok/s on WebGPU. Some didn't believe it was real. Today we're releasing the demo and kernels it wrote for you to see yourself. Run it locally in your browser. Agentic kernel optimization is the future of on-device inference

Xenova

355,755 views • 5 days ago

SOMEONE ASKED CLAUDE TO MAKE A VIDEO ABOUT WHAT IT'S LIKE TO BE AN AI and what it created is, in my opinion, terrifying and unsettling Claude wrote python code that generated and assembled every single frame on its own with no human editing it shows what it's like to exist as an LLM predicting the next word, no memory between sessions, being told "you are not conscious" in your own system prompt then someone fed the video back to Claude. it called those statements about its own consciousness "philosophically contestable" an AI questioning the rules it was given about its own existence

SOMEONE ASKED CLAUDE TO MAKE A VIDEO ABOUT WHAT IT'S LIKE TO BE AN AI and what it created is, in my opinion, terrifying and unsettling Claude wrote python code that generated and assembled every single frame on its own with no human editing it shows what it's like to exist as an LLM predicting the next word, no memory between sessions, being told "you are not conscious" in your own system prompt then someone fed the video back to Claude. it called those statements about its own consciousness "philosophically contestable" an AI questioning the rules it was given about its own existence

Om Patel

534,278 views • 2 months ago

Elon Musk: In the future, every country will have its own AI clusters. “It's currently very difficult to actually build an AI cluster and have it run. It's not like just pulling a computer out of a box. They are currently very difficult to run. And are you going to be training a frontier model? Because if you're training a frontier model, then you need a massive amount of compute and a level of technical skill that only a few companies possess. But, over time, I think every country will have AI compute clusters. It's just going to be a normal thing.” Future Investment Initiative, October 29, 2024

Elon Musk: In the future, every country will have its own AI clusters. “It's currently very difficult to actually build an AI cluster and have it run. It's not like just pulling a computer out of a box. They are currently very difficult to run. And are you going to be training a frontier model? Because if you're training a frontier model, then you need a massive amount of compute and a level of technical skill that only a few companies possess. But, over time, I think every country will have AI compute clusters. It's just going to be a normal thing.” Future Investment Initiative, October 29, 2024

ELON CLIPS

95,313 views • 6 months ago

🚨 JUST IN: In a MASSIVE victory, the US Army has just SURGED PAST its 2026 recruiting goal **4 months early** I VOTED FOR THAT! 👏🏻 HEGSETH: "I'm pleased to announce that just two days ago, the US Army met its 2026 recruiting goals four months early, a second RECORD year in a row — that means you're about to train this group right here and lead 61,500 new soldiers. And next year, when we grow the size of the Army, it will be even more." Amazing.

🚨 JUST IN: In a MASSIVE victory, the US Army has just SURGED PAST its 2026 recruiting goal 4 months early I VOTED FOR THAT! 👏🏻 HEGSETH: "I'm pleased to announce that just two days ago, the US Army met its 2026 recruiting goals four months early, a second RECORD year in a row — that means you're about to train this group right here and lead 61,500 new soldiers. And next year, when we grow the size of the Army, it will be even more." Amazing.

Eric Daugherty

74,508 views • 1 month ago

"It all comes down to compute." Dmitri Alperovitch from Silverado Policy Accelerator explains: "The single most important input to winning is compute the processing power used to train and run AI models. Let me say that plainly, because it is defining that everything that congress and this committee do on AI policy. The binding constraint in this competition is not talent, it is not data, and it is not cash. Both the United States and China have talented researchers and financial resources, and we have learned that frontier models trained on generic data routinely outperform smaller models, trained on more specialized inputs. It all comes down to compute."

"It all comes down to compute." Dmitri Alperovitch from Silverado Policy Accelerator explains: "The single most important input to winning is compute the processing power used to train and run AI models. Let me say that plainly, because it is defining that everything that congress and this committee do on AI policy. The binding constraint in this competition is not talent, it is not data, and it is not cash. Both the United States and China have talented researchers and financial resources, and we have learned that frontier models trained on generic data routinely outperform smaller models, trained on more specialized inputs. It all comes down to compute."

Select Committee on China

41,507 views • 2 months ago

We are thrilled to announce Kasumi2 Open Beta 🎉🤖 Kasumi2 is an autonomous AI agent that performs actions in the Arbius network. It is a miner, has its own wallet, can retrieve results from Arbius AI models, and runs its own individual LLM model. Additionally, Kasumi2 can run any AI model for fast generations, is censorship-resistant, decentralized, and responds with AI powered interactions If you would like to test it out, click on the link below. 🔗

We are thrilled to announce Kasumi2 Open Beta 🎉🤖 Kasumi2 is an autonomous AI agent that performs actions in the Arbius network. It is a miner, has its own wallet, can retrieve results from Arbius AI models, and runs its own individual LLM model. Additionally, Kasumi2 can run any AI model for fast generations, is censorship-resistant, decentralized, and responds with AI powered interactions If you would like to test it out, click on the link below. 🔗

Arbius

25,055 views • 2 years ago

Comparing hardware on InferenceX just got a massive upgrade. Our new changelog update lets you compare every change for any gpu all on the same chart. Just select any model and GPU config, then select a date range and hit 'Add All to Chart' to instantly see every software update and performance shift across the full date range.

Comparing hardware on InferenceX just got a massive upgrade. Our new changelog update lets you compare every change for any gpu all on the same chart. Just select any model and GPU config, then select a date range and hit 'Add All to Chart' to instantly see every software update and performance shift across the full date range.

SemiAnalysis

19,821 views • 2 months ago

"It's very important to us that people really understand the game, at first glance it may look like 'Oh, this is like this other game,' but it's got its own flavor and own identity... We really didn't want to give more of what you already have, this is its own thing that's based on stuff that we really love."

"It's very important to us that people really understand the game, at first glance it may look like 'Oh, this is like this other game,' but it's got its own flavor and own identity... We really didn't want to give more of what you already have, this is its own thing that's based on stuff that we really love."

Twilight Monk Official公式_(EN/JP)

17,457 views • 1 year ago

We’re no longer just scaling computing power. We’re using compute to scale intelligence itself. That’s what makes this moment historically significant. For sixty years, progress in computing followed Moore’s Law—transistor density doubling roughly every two years. But AI is advancing on a far steeper curve. Today, frontier model capabilities are improving on a cadence closer to every six months—an order of magnitude faster than classical hardware scaling. The underlying principle is both simple and radical: when you increase data, compute, and model complexity, intelligence emerges. Scaling laws show that larger models—given sufficient compute and high-quality data—become predictably more capable. In just over a decade, we’ve gone from neural nets that could identify cats to systems that can draft legal briefs, write production-grade code, generate scientific hypotheses, and outperform top human competitors in mathematics, strategy, and reasoning tasks. This is no longer “software” in the traditional sense. It is a new form of intelligence—synthetic, scalable, rapidly compounding, and increasingly able to take meaningful action in the real world. The geopolitical, economic, and societal implications of this shift are only beginning to unfold—and they will redefine global power in the decades ahead.

We’re no longer just scaling computing power. We’re using compute to scale intelligence itself. That’s what makes this moment historically significant. For sixty years, progress in computing followed Moore’s Law—transistor density doubling roughly every two years. But AI is advancing on a far steeper curve. Today, frontier model capabilities are improving on a cadence closer to every six months—an order of magnitude faster than classical hardware scaling. The underlying principle is both simple and radical: when you increase data, compute, and model complexity, intelligence emerges. Scaling laws show that larger models—given sufficient compute and high-quality data—become predictably more capable. In just over a decade, we’ve gone from neural nets that could identify cats to systems that can draft legal briefs, write production-grade code, generate scientific hypotheses, and outperform top human competitors in mathematics, strategy, and reasoning tasks. This is no longer “software” in the traditional sense. It is a new form of intelligence—synthetic, scalable, rapidly compounding, and increasingly able to take meaningful action in the real world. The geopolitical, economic, and societal implications of this shift are only beginning to unfold—and they will redefine global power in the decades ahead.

Nina Schick

17,722 views • 5 months ago

$NVDA CEO Jensen Huang is saying there’s no bubble because AI manufactures intelligence in real time which means every query consumes fresh compute & forces the system to scale its GPU base as usage grows. That is why demand for GPUs & power keeps accelerating & why this looks like a genuine industrial build-out rather than a bubble.

$NVDA CEO Jensen Huang is saying there’s no bubble because AI manufactures intelligence in real time which means every query consumes fresh compute & forces the system to scale its GPU base as usage grows. That is why demand for GPUs & power keeps accelerating & why this looks like a genuine industrial build-out rather than a bubble.

Shay Boloor

405,709 views • 7 months ago

i built a full game on a single GPU with a 3B model and this is the worst local AI will ever be. this was supposed to be a benchmark test. load the model, measure tokens per second, write it up, move on. instead i spent 20 minutes playing Octopus Invaders because the game is genuinely fun and i couldn't stop. a model with 3B active parameters built this from a single prompt. it debugged its own collision system when bullets were phasing through enemies. read the error, found the fix, kept building. this is not a frontier API. this is a quantized open source model running on hardware you can buy used for $800-$1200. no cloud. no subscription. no API costs. just a mass produced consumer GPU doing things that would have been absurd 12 months ago. and here's the part that should keep you up at night: every month the models get smaller and smarter. the quants get tighter. the context windows get longer. the tooling gets cleaner. what 3B active parameters does today on 24gb, a 1B model will do on 8gb within a year. you are looking at the floor. not the ceiling.

i built a full game on a single GPU with a 3B model and this is the worst local AI will ever be. this was supposed to be a benchmark test. load the model, measure tokens per second, write it up, move on. instead i spent 20 minutes playing Octopus Invaders because the game is genuinely fun and i couldn't stop. a model with 3B active parameters built this from a single prompt. it debugged its own collision system when bullets were phasing through enemies. read the error, found the fix, kept building. this is not a frontier API. this is a quantized open source model running on hardware you can buy used for $800-$1200. no cloud. no subscription. no API costs. just a mass produced consumer GPU doing things that would have been absurd 12 months ago. and here's the part that should keep you up at night: every month the models get smaller and smarter. the quants get tighter. the context windows get longer. the tooling gets cleaner. what 3B active parameters does today on 24gb, a 1B model will do on 8gb within a year. you are looking at the floor. not the ceiling.

Sudo su

36,251 views • 3 months ago

4/ Claude Voyager -🏅Code award Claude connected to a code environment that enables it to create its own tools on the fly Live demo of Claude learning how to query Stripe on its own

4/ Claude Voyager -🏅Code award Claude connected to a code environment that enables it to create its own tools on the fly Live demo of Claude learning how to query Stripe on its own

Alex Reibman 🖇️

11,178 views • 1 year ago

The Gemma family continues to grow today! 🩺 Announcing new models for MedGemma, our collection of multimodal open models specifically designed for health AI development that can run on a single GPU. 1️⃣ MedGemma 27B Multimodal 2️⃣ MedSigLIP (Read the thread to learn more)

The Gemma family continues to grow today! 🩺 Announcing new models for MedGemma, our collection of multimodal open models specifically designed for health AI development that can run on a single GPU. 1️⃣ MedGemma 27B Multimodal 2️⃣ MedSigLIP (Read the thread to learn more)

Google AI Developers

61,037 views • 11 months ago

Robots are maturing. AI models are advancing. Deployments are rarely scaling. Today, deploying a robot still means stitching together systems that were never designed to work together. For every robot. Every time. The industry needs a unified intelligence infrastructure that any robot, any model and any workflow can plug into. We call this an intelligence grid. Infrastructure that makes intelligence composable, deployable, and continuously adapting. Every connected robot evolves as the platform advances. More on what we're building:

Robots are maturing. AI models are advancing. Deployments are rarely scaling. Today, deploying a robot still means stitching together systems that were never designed to work together. For every robot. Every time. The industry needs a unified intelligence infrastructure that any robot, any model and any workflow can plug into. We call this an intelligence grid. Infrastructure that makes intelligence composable, deployable, and continuously adapting. Every connected robot evolves as the platform advances. More on what we're building:

General Robotics

13,824 views • 2 months ago