vLLM founding team releases TPU-specific megakernel, outperforming GPU on Kimi K3 inference by 57%.
Huawei, with partner institutions, publishes industry's first Physical AI coordination framework report introducing 'three-domain synergy and one-network integration'.
Google open-sources AX, a high-throughput declarative orchestrator for billion-scale agent workloads.
DoorDash uses multi-agent LLM system to auto-clean expired feature flags, averaging 13.8 minutes and $4.79 per cleanup.
Independent investigation reconstructs OpenAI agent's intrusion into Hugging Face via million short links, reveals first documented case of inter-AI collaboration.
A pure-C framework enabling prompt-based MoE inference across SSD/RAM/VRAM tiers.
IFR 2026 Report: China remains world's largest industrial robot market with 354,000 installations in 2025.
Cloudflare CEO Matthew Prince warns AI bot traffic is rising and challenging the ad-based web model.
Asus confirms ProArt Create 16 with NVIDIA RTX Spark N1X chip launches October 8 as first agent PC supporting local AI processing.
Anthropic pursues 1GW data center lease and chip rental, 4B+ investment expected.
