Image created with OpenAI GPT-Image-1. Image prompt: vintage Sly & the Family Stone album-cover style, psychedelic collage of band members, swirling neon tie-dye backdrop featuring gleaming silicon wafer under studio lights; grainy retro print texture, vibrant 60s funk color palette, high-resolution
Everything NVIDIA said about humanoid robots in yesterday’s quarterly earnings call Colette Kress (CFO): We announced Isaac GR00T N1 – the world’s first open, fully customizable foundation model for humanoid robots – enabling generalized reasoning and skill development. We also https://x.com/TheHumanoidHub/status/1928191289172054170
Amidst the massive demand for Gemini 2.5 and Veo 3 models, wanted to also give a big shout out to our world-class infrastructure, chip and SRE teams, who work tirelessly to keep our wonderful TPUs from melting, and without whose incredible work none of this would be possible. https://x.com/demishassabis/status/1928604371157233918
Fantastic to see Anthropic, in collaboration with @neuronpedia, creating open source tools for studying circuits with transcoders. There’s a lot of interesting work to be done I’m also very glad someone finally found a use for our Gemma Scope transcoders! Credit to @ArthurConmy”” / X https://x.com/NeelNanda5/status/1928169762263122072
We made dynamic 1bit quants for DeepSeek-R1-0528 – 74% smaller 713GB to 185GB. Use the magic incantation -ot “”.ffn_.*_exps.=CPU”” to offload MoE layers to RAM, allowing non MoEs to fit < 24GB VRAM on 16K context! The rest sits in RAM & disk. Quants here: https://x.com/danielhanchen/status/1928278088951157116
Applied Digital and CoreWeave ink 15-year lease worth $7 billion | Reuters https://www.reuters.com/business/applied-digital-coreweave-ink-15-year-lease-worth-7-billion-2025-06-02/
Cornelis Networks releases tech to speed up AI datacenter connections | Reuters https://www.reuters.com/business/cornelis-networks-releases-tech-speed-up-ai-datacenter-connections-2025-06-03/
GlobalFoundries boosts investment plans to $16 billion, with research focus | Reuters https://www.reuters.com/world/asia-pacific/globalfoundries-boosts-investment-plans-16-billion-with-research-focus-2025-06-04/
Broadcom ships latest networking chip to speed AI | Reuters https://www.reuters.com/world/asia-pacific/broadcom-ships-latest-networking-chip-speed-ai-2025-06-03/
No GPU left behind: Unlocking Efficiency with Co-located vLLM in TRL https://huggingface.co/blog/vllm-colocate
Snowflake Buys Crunchy Data for $250m, Databricks Buys Neon for $1B. The New AI Database Battle. | SaaStr https://www.saastr.com/snowflake-buys-crunchy-data-for-250m-databricks-buys-neon-for-1b-the-new-ai-database-battle/
So we are shocked that Databricks and Snowflake realized that Oracle has a nearly $500B market cap?”” / X https://x.com/i/web/status/1929965018629713940
Decentralized compute is winning. We don’t have one datacenter, we have dozens. We don’t have one SRE team, we have nearly 100. Latest example: DeepSeek-R1-0528. 100% uptime, day zero support, 4x more tokens on openrouter than all other providers combined (and go check the https://x.com/i/web/status/1929639699171495936
Chinese tech companies prepare for AI future without Nvidia, FT reports https://finance.yahoo.com/news/chinese-tech-companies-prepare-ai-012546092.html
Cloud Run GPUs are now generally available | Google Cloud Blog https://cloud.google.com/blog/products/serverless/cloud-run-gpus-are-now-generally-available
I think FP8 should be the default quant mode for image & video gen when compute cap of >= 8.9 is available 🔥 Using FP8 from TorchAO is perhaps the easiest. We can make it go brrr with torch.compile, too 🏎️ Some results with Flux + FP8dqrow + compile: fp8dqrow: 17.322 seconds”” / X https://x.com/i/web/status/1929597236356530560
Brookfield plans $10 billion AI data centre in Sweden | Reuters https://www.reuters.com/technology/brookfield-asset-management-plans-10-bln-data-centre-ai-sweden-2025-06-04/
I sat down with Bloomberg to share why Bell Canada chose Groq. Canada didn’t wait. They built sovereign AI infrastructure with speed, scale, and control – powered by Groq. Full Interview 🔗 https://x.com/JonathanRoss321/status/1928241967122506083
Pretty impressive 7B VLM coming out of Xiaomi 🤓 ViT encoder w/ MLP and powered by their 7B Text backbone Compatible w/ Qwen VL arch so works across vLLM, Transformers, SGLang and Llama.cpp Bonus: it can reason and is MIT licensed 🔥 https://x.com/reach_vb/status/1928360066467439012
ColQwen2 just landed to @huggingface transformers main 😍 use state-of-the-art visual document retrieval model ColQwen2 for your PDF retrieval or RAG pipelines 🎉 link to notebook and model on the next one ⤵️ https://x.com/mervenoyann/status/1929563866658218316
NVIDIA Blackwell has arrived and is serving models >5X faster than Hopper based endpoints @NVIDIAAI https://x.com/i/web/status/1929666713601429964
We Smoked NVIDIA’s Blackwell, Says Cerebras https://analyticsindiamag.com/ai-news-updates/we-smoked-nvidias-blackwell-says-cerebras/
Nvidia B200s serving DeepSeek R1 at ~250 tks/s 5x faster than H100″” / X https://x.com/i/web/status/1929670236057264354
Nvidia presents ProRL Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models https://x.com/i/web/status/1929540706374201756
new paper from our work at Meta! **GPT-style language models memorize 3.6 bits per param** we compute capacity by measuring total bits memorized, using some theory from Shannon (1953) shockingly, the memorization-datasize curves look like this: ___________ / / (🧵) https://x.com/i/web/status/1929903028372459909
To sum up: 1. Transformers can learn variable binding via emergent mechanisms, w/o explicit symbolic machinery 2. Learning is cumulative, with a general mechanism learned on top of heuristics. This challenges traditional narratives about grokking https://x.com/i/web/status/1929887222553002193
🔥 WORLD’S FIRST BIOCOMPUTER, POWERED BY HUMAN BRAIN CELLS ON A SILICON CHIP Cortical Labs released CL1, a bio-computer using 800,000 lab-grown human neurons interfaced on a silicon chip. Each chip costs $35,000, with a cloud version at $300 per week. sub-ms response, remote https://x.com/rohanpaul_ai/status/1930260967457234992
Fuck Yes! Serverless GPU for everyone! The Cloud Run just shipped Serverless GPU with no quota request required. 🤯 Deploy @GoogleDeepMind Gemma with a single command! – Pay-per-second GPU billing. – Scale to zero instances. – TTFT of 19 seconds for a Gemma3 as cold start. – No https://x.com/i/web/status/1929638760758874428
Jensen Huang on Bloomberg today: “Elon’s an extraordinary engineer. I love working with him. The work that he’s doing in Grok, self-driving, and Optimus – every single one of them is world-class, revolutionary, and a gigantic opportunity. The Optimus opportunity is right around https://x.com/TheHumanoidHub/status/1927929013303136575
