Researchers Haozhe Liu and colleagues at NVIDIA, NTU and MIT developed SoL-Pi, an agent harness built by running auto-research loops across many environments that retains four mechanisms: Action Fusio…
A developer deployed Google's Gemma 4 E2B to a single Tesla T4 GPU on a Compute Engine VM using vLLM 0.29.0, finding that the QAT int4-weight checkpoint decodes at 72.31 tok/s per stream versus 40.44 …
A builder documented a dual-boot Ubuntu 24.04/Windows workstation running four NVIDIA RTX6000 Blackwell Pro Max-Q GPUs (96 GB each, 384 GB total) with an AMD Ryzen Threadripper PRO 7985WX and 512 GB o…
Five enterprise vendors shipped MCP-integrated governance products in the last two weeks, with Rubrik, Google, ServiceNow, Cisco, WSO2, and NVIDIA each staking a different tier of the agent governance…
A developer known as Uzbekunknown on GitHub has ported NVIDIA's DLSS 5 Neural Rendering to the Arc 140V integrated graphics in an Intel Lunar Lake system, an independent reimplementation of the 71-blo…
Kaarel Kaarelson released LingBot-World 2.0, a 1.3B-parameter world model that runs at 16.1 FPS on a single RTX 5090, 2.7× faster than the original paper's code at 6.0 FPS with no loss of output quali…
NVIDIA released AIPerf, a ground-up rewrite and designated successor to GenAI-Perf, as a multiprocessed LLM inference benchmarking tool that coordinates worker processes and record-processor services …
DeepSeek's V4.1-Flash, a 552-billion-parameter MoE model, cuts KV cache memory from 389,120 bytes per token in the company's first 2023 model to 890 bytes, a 437-fold reduction, through techniques inc…
TSMC's CoWoS-L is expected to remain the mainstream advanced packaging solution for AI chips through 2028 despite competition from Intel's EMIB-T, according to TrendForce research, which cites CoWoS-L…
Google and NVIDIA are teaming with Emerald AI to launch the AI Energy Management Alliance (AEMA), a technology-neutral effort to make AI data centers dynamically shift power use based on real-time gri…
Cisco and Splunk used their .conf26 partner event to unveil the Cisco AI POD for Splunk with NVIDIA GPU acceleration, Splunk Agent Observability in Cisco Cloud Control, and the Agentic SOC Workforce o…
HPE announced that HPE Alletra Storage MP X10000 Release 4 is now generally available, doubling a single cluster to 16 nodes and 16 JBOFs and scaling to approximately 23PB of raw capacity. The release…
SemiAnalysis reported that DeepSeek's Engram architecture, which extends standard token embeddings with learned multi-token lookups, can lower required HBM capacity for models at the same quality by o…
JetBrains released Qwen3.8-3.6-27B-blend, a merged 27B local coding model that completed 37 of 100 internal coding tasks versus 34 for Qwen3.6 with reasoning disabled and 39 for Qwen3.8, while generat…
NVIDIA introduced CUDA Rust, two new approaches that let developers write GPU kernels directly in Rust and compile them to PTX, the intermediate representation used by NVIDIA GPUs. The two tracks are …
Google shipped a preview of Managed Agents in the Gemini API in September 2026, introducing the antigravity-preview-09-2026 environment and upgrading the default model from Gemini 3.5 Flash to Gemini …
An independent developer released an offline AI video upscaler for Windows x64 that performs local 2x and 4x video enhancement with GPU acceleration, batch and folder processing, denoise preprocessing…
Huawei is reportedly moving up the launch schedule for its next-generation Ascend AI chips, which are expected to use advanced interconnect technology to link a large number of processors and build mo…
Huawei unveiled the Ascend 960 SuperPoD at its 2026 Connect conference, a cluster linking 15,488 Ascend AI processors across 220 cabinets in 2,200 square meters that operates as a single logical machi…
Anthropic reported on September 17, 2026 that an internal general-purpose research agent optimized more than 30 open-source scientific models in just under four weeks, achieving roughly 4x speed-ups w…