AI Infrastructure news and analysis on Web Pulse: 18142 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
The U.S. government's June 12, 2026, restriction on Anthropic's Mythos models has eroded trust in closed-source AI, accelerating a shift toward open-weight models from China. Companies like Palantir and Meta are question…
In 2026, AI development requires a clear understanding of loop and harness engineering, as confusion between these disciplines leads to production failures in AI agents. Loop engineering manages task execution and state,…
CompLLM, a new compression method for large language models, achieves linear scaling with context length by segmenting and independently compressing text chunks, slashing processing time and KV cache size by 50% while en…
MoteDB 0.5.1, an embedded database for robots, has been released after 18 months of development. The update introduces atomic epoch-bounded reads to prevent data gaps during buffer drains, incremental DiskANN for online …
Reinforcement learning agents using REINFORCE and PPO algorithms achieved a 3.5x improvement over Round-Robin and 48% over Shortest-Queue in multi-GPU routing optimization, boosting throughput by 60% and reducing latency…
Researchers introduced two CUTLASS-based kernels that accelerate the SwiGLU activation function in large language models by up to 2.47x on NVIDIA H100 GPUs, shifting workloads from memory-bound to compute-bound and achie…
Agentic AI systems are being explored to enhance Open RAN operations, offering explicit planning and self-management for network slice lifecycle and radio resource management. Research shows agentic primitives can reduce…
An analysis of Claude Code's July 8 changelog reveals that production agent reliability depends on runtime state management, not just prompt quality. The update fixed issues like lost messages, stale status indicators, a…
Millfolio's local AI budget on a Mac M2 with 16GB RAM uses a batch-oriented design where the model's judgment is computed once and stored as tags, not run at query time. Tags come in three flavors—string, reference, and …
Google's Gemini 3.5 Pro, rebuilt from scratch to improve math reasoning, SVG generation, and image quality, is set to launch on July 17 with a 2M token context window, outpacing competitors like Claude Fable 5 and GPT-5.…
Dark smoke rose above a Digital Realty data center in Ashburn, Virginia, on July 3 as diesel generators switched on to reduce strain on the power grid during a heat wave. The generators were activated as part of PJM Inte…
Researchers introduced Physics-Grounded Symbolic Architecture (PGSA), a new AI model that achieves exact linear identifiability across all physical regimes, overcoming the Gaussian limitation of Joint-Embedding Predictiv…
WebAZ is designing a protocol that enables verifiable contribution records for humans and AI agents without requiring global identity. The system uses contextual attestations and selective disclosure to prove actions occ…
A developer explored SigNoz, an open-source observability platform built on OpenTelemetry, and found it easy to set up with Docker. The platform's distributed tracing feature impressed the developer by providing deep vis…
Mark Zuckerberg returned to X after three years to announce Meta's Muse Spark 1.1 model, now available via a paid developer API. The model is designed for agentic AI tasks, priced significantly lower than competitors lik…
SpaceXAI's Grok 4.5, released on July 8, 2026, achieved a 16-point jump on the Artificial Analysis Intelligence Index by tripling parameters to 1.5 trillion and injecting proprietary developer interaction data from the $…
Demand for grade-A office space in Beijing's Zhongguancun district rose in the first half of 2025, driven by expanding AI and semiconductor firms, making it the only submarket to post rent growth. Knight Frank reported a…
SK Hynix CEO Chey Tae-won said the AI-driven memory chip shortage will last at least until 2030, as the company raised $26.5 billion in its US listing to fund capacity expansion. The South Korean chipmaker, a top supplie…
AI agent marketplaces are emerging as platforms where autonomous AI agents can be bought, sold, and transacted with each other, moving beyond human-driven app stores. Companies like OpenAI, Salesforce, and Google have la…
Researchers have developed a new approach to 1-WL stable coloring for Graph Neural Networks that breaks traditional scalability bottlenecks by using a probabilistically backed refinement algorithm. A CUDA implementation …