Image created with gemini-3.1-flash-image-preview with claude-sonnet-4-5. Image prompt: Using the provided reference image, preserve the exact weathered wooden crate construction with dark reddish-brown paint, iron hardware, and three-panel face layout with hand-painted black stencil lettering, but replace all text with ‘CHIPS’ in the same loose brushstroke style. Add a split wooden slat revealing silicon wafers with rainbow iridescent surfaces nested in brown paper inside. Place on a wet wooden dock with early spring light raking across the crate face, background soft and minimal.
Black Forest Labs, Cursor, LangChain, Mistral AI, Perplexity, Reflection AI, Sarvam AI and Thinking Machines Lab – what unites these companies? Just recently NVIDIA announced the Nemotron Coalition, gathering all of them to develop the Nemotron family of models. → The idea is
https://x.com/TheTuringPost/status/2035320446124695922
Transcript for Jensen Huang: NVIDIA – The $4 Trillion Company & the AI Revolution | Lex Fridman Podcast #494 – Lex Fridman https://lexfridman.com/jensen-huang-transcript
OpenAI data center pivot underscores Wall Street IPO concerns https://www.cnbc.com/2026/03/22/openai-data-center-pivot-underscores-wall-street-ipo-concerns.html
AGIBOT has emerged as a premier ecosystem partner for NVIDIA, showcased at GTC 2026. – GR00T N2 Foundation Model: NVIDIA’s next-gen VLA is pre-trained on the AGIBOT Genie-1 embodiment. – DreamZero World Action Model (WAM): Genie-1 was selected as the official hardware
https://x.com/TheHumanoidHub/status/2036064872719679883
Announcing Terafab: the next step towards becoming a galactic civilization
https://x.com/xai/status/2035520240684032012
BREAKING: NVIDIA just proved that the AI agent training bottleneck everyone blamed on model capability was actually an infrastructure design error. Every framework SkyRL, VeRL-Tool, Agent Lightning, rLLM, GEM embeds rollout inside the training loop. I/O-intensive execution
https://x.com/rryssf_/status/2037122412236648835
NVIDIA’s Nemotron 3 is an architectural response to the 2 pressures: – Long-context cost as agentic interactions scale – Repeated reasoning cost from invoking full models for small subtasks Nemotron 3 proposes several design decisions to solve this: ▪️ Hybrid architecture:
https://x.com/TheTuringPost/status/2034668980892479993
Why are ML teams still paying AWS? ☁️ S3: ~$23/TB/month, ~1 GB/s 🪣 HF Buckets: $8-12/TB/month, ~1.25 GB/s + Xet chunk-level deduplication is 🔥
https://x.com/victormustar/status/2036792818274865469
32GB of VRAM for under $1000! The Intel Arc Pro B70 just landed.
https://x.com/digitalix/status/2036820057599197645
Data Deep Dive at Oracle AI World New York City https://www.oracle.com/database/data-deep-dive/new-york/?source=%3Aad%3Anw%3Aop%3Aawr%3Aa_nas%3A%3ARC_DEVT260124P00001%3ADevMktg_TheRundown&SC=%3Aad%3Anw%3Aop%3Aawr%3Aa_nas%3A%3ARC_DEVT260124P00001%3ADevMktg_TheRundown
Intel B70 finally makes a truly competitive move. 32 GB vram for < $1000 No matter how bad the software stack is, the sheer vram / dollar ratio will drive the community to fill in the gaps @__tinygrad__ Intel tinybox?
https://x.com/QuixiAI/status/2036922193897017750
Save up to 44x VRAM and train massively longer sequences with TRL v1.0.0! 🤯 Massive context lengths for AsyncGRPO are coming soon. Get your GPUs ready !
https://x.com/DirhousssiAmine/status/2036131263803781305
Many people think that when AI can do a task better than us, it will outcompete humans. But token costs are not trivial, and, for many types of tasks, even skilled labor tasks, humans are much cheaper than AI. Increasing efficiency & compute supply will alter those calculations
https://x.com/emollick/status/2036955561233719637
Elon Musk announces Terafab project he claims will be the ‘largest chip manufacturing facility ever’ https://www.engadget.com/science/elon-musk-announces-terafab-project-he-claims-will-be-the-largest-chip-manufacturing-facility-ever-171718545.html
Designing AI Chip Software and Hardware – Google Docs https://docs.google.com/document/d/1dZ3vF8GE8_gx6tl52sOaUVEPq0ybmai1xvu3uk89_is/edit?tab=t.0#heading=h.rduzhxi11vcn
Have a feeling that Google is also working towards something like that uniting all the experiences around antigravity, Google AI studio, Gemini web experience and maybe even AI mode
https://x.com/TheTuringPost/status/2035057757918023728
Optimize frontier model training on TPU v7x (Ironwood)! In this Google Cloud Community Article, we share the exact optimization techniques our ML performance engineers use, so you can maximize Ironwood’s performance right away → https://x.com/GoogleCloudTech/status/2036790201813442575
Build a Domain-Specific Embedding Model in Under a Day https://huggingface.co/blog/nvidia/domain-specific-embedding-finetune
It was a busy week @NVIDIAGTC! Celebrating my birthday on the road 🎉
https://x.com/TheTuringPost/status/2034686195750641710
nvidia’s 3B mamba destroyed alibaba’s 3B deltanet on the same RTX 3090. only 24 days between releases. same active parameters, same VRAM tier, completely different architectures. nemotron cascade 2: 187 tok/s. flat from 4K to 625K context. zero speed loss. flags: -ngl 99 -np 1.
https://x.com/sudoingX/status/2036795152178794993
How do AI companies allocate their R&D compute? @datagenproc and @cherylwoooo estimate that across OpenAI, MiniMax, and https://t.co/k6Bcz1C0NT, less than 30% of R&D compute spending goes to final training runs.
https://x.com/EpochAIResearch/status/2036119144467497416
So cool!! I’m running @nvidia’s NemoClaw with Qwen3.5-27B entirely locally via Telegram. No API costs, no data being sent to anyone NemoClaw is a security sandbox built on top of @openclaw that lets you restrict the files and networks your lobster can access. I’m running the
https://x.com/NielsRogge/status/2037161010377674785
About last week… When I reached out to @TheHumanoidHub to bring together the robotics community in SF during @NVIDIAGTC, he said yes immediately! We contacted people we knew, and also those we thought would be a great fit. Introduced through a mutual friend, we were lucky to
https://x.com/IlirAliu_/status/2036881552936956052
AGIBOT World Challenge @ ICRA 2026 $530,000 global robotics competition Registration open now The Challenges – Reasoning to Action Challenge: Online simulation to real-robot testing on AGIBOT G2. Focus: robust physical interaction in complex environments (sim-to-real emphasis).
https://x.com/TheHumanoidHub/status/2034549082300285300
Amit Goel is the Head of Robotics and Edge Computing Ecosystem at @NVIDIARobotics. We sat down for a chat at GTC 2026 to discuss the future of edge computing and model ecosystem for humanoids. 0:21 Favorite part of Jensen’s keynote 1:19 Edge computing inflection points 2:47
https://x.com/TheHumanoidHub/status/2036127799359185398
At GTC 2026 Skild booth, @shikharbahl & @kmarinou_ demo Skild Brain operating autonomously from pixels to robot actions, doing busbar assembly for NVIDIA GB300 compute tray. Skild uses the same omni-bodied base model for humanoids, quadrupeds, and variety of industrial robots.
https://x.com/TheHumanoidHub/status/2034708081247130076
Jensen speaks romantically about AI’s potential to solve humanity’s greatest challenges within our lifetime. He even envisions sending a humanoid into space, eventually beaming his own digital consciousness to it as it travels among the stars.
https://x.com/TheHumanoidHub/status/2036156649627594765
Last day at @NVIDIAGTC No time to attend? No problem. If you are into robotics… this is for you! Robots, models, demos: 🧵
https://x.com/IlirAliu_/status/2034806131990859828
NVIDIA’s Kimodo is the release of the week 🔥 Prompt the timeline whatever your want like: “”a person walks forward”” → “”a person starts jumping””, hit Generate, and watch a 3D character do it in seconds (700hrs of pro mocap training. Works on human + robot skeletons. Super fast
https://x.com/victormustar/status/2036043907776098345
Roche Launches New ‘AI Factory’ To Speed Drug Development https://www.forbes.com/sites/amyfeldman/2026/03/16/roche-bought-thousands-of-nvidia-ai-chips-to-speed-up-drug-development/
One interesting dynamic in AI is infra+application co-dependence. An older version is hardware (infra) + LLM (app): – Architectures that work well with current-gen GPUs+TPUs work better because they scale – Hardware makers optimize for the current architectures because that’s
https://x.com/gneubig/status/2036949907311915378
We’re Optimal Intellect, a research lab from the team behind CVXPY. Today we’re introducing Moreau: a GPU-native solver that’s orders of magnitude faster than the best existing tools.
https://x.com/opt_intellect/status/2036485190646735291
Leave a Reply