100% stood the test of time: “What Ilya saw” was Sam’s bad behavior, not AGI.
AI HARDWARE
-
Humanoid Reality: Nascent Market Faces Quality Challenges, Improvement Expected
By
–
The humanoid reality – nascent and emerging market with product quality challenges. Expect that to change over the next 5 to 10 years.
-
Maintaining Training and Inference Capacity for AI Frontier
By
–
You don’t give up training or inference capacity if you are trying to stay on the frontier…
-
Ouster’s Native Color SPAD Pixels for Visible Light Capture
By
–
we need more benchmarks! awesome work by harvey here, and excited to work with them
-
NVIDIA AI announces TokenSpeed, a fast inference engine for agentic workloads
By
–
TokenSpeed is a brand new inference engine purpose built for speed-of-light agentic workloads. Read their blog to learn more about its advanced KV cache management, safe and efficient scheduler, and pluggable layered kernel system designed for multi-silicon support. Plus, it
-
Anthropic Partners with SpaceX for Supercluster Capacity, Doubling Rate Limits
By
–
Anthropic is partnering with SpaceX to use the capacity in their Colossus supercluster! Doubling rate limits!
-
Gigafactory Scale Cryogenic Regeneration and HTS-MHD Synergies
By
–
confused; wasn’t the point that they would not be able to solve it via memorization?
-
Perplexity develops ROSE inference engine with CuTeDSL for faster GPU kernels
By
–
We’ve developed our own inference engine Runtime-Optimized Serving Engine (ROSE) to serve models ranging from embeddings to trillion-parameter LLMs. With CuTeDSL integrated into our inference engine, Perplexity can build the specialized GPU kernels faster to bring models up to
-
Evolution of AI Chips and Computing Architectures
By
–
AI chips didn’t just appear overnight. We went from general-purpose CPUs → big data → ML driving new architectures entirely.
— SambaNova (@SambaNovaAI) 6 mai 2026
Now it’s not one path forward, it’s different approaches to compute & efficiency.
Great @dcdnews EP with @SumtiJairath.
🎧 Listen here:… pic.twitter.com/OIwUEK5wRuAI chips didn’t just appear overnight. We went from general-purpose CPUs → big data → ML driving new architectures entirely. Now it’s not one path forward, it’s different approaches to compute & efficiency. Great @dcdnews EP with @SumtiJairath
. Listen here: -
NVIDIA Open-Sources MRC Transport Protocol for AI Training Clusters
By
–
NVIDIA just open-sourced a transport protocol that powers OpenAI's Blackwell clusters. It opened MRC, a new RDMA transport protocol for massive AI training clusters. Instead of pushing GPU traffic through one fragile path, MRC spreads a single connection across multiple network