AI Infrastructure news and analysis on Web Pulse: 35108 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
A developer's technical blog series on building a secure AI platform details the backend foundation, emphasizing a deterministic startup lifecycle, explicit configuration validation, and layered architecture to prevent i…
Runtime enforcement is crystallizing as a mandatory infrastructure layer for AI agents, with ServiceNow, Microsoft, MintMCP, Citrix, and CrowdStrike converging on a three-layer stack of connectivity, security/governance,…
Cobrainer, a Munich-based skills-intelligence company, has built a graph-based agent memory system on SurrealDB, avoiding the need for separate vector, graph, or search engines. The system enables an AI agent to automati…
AMD's experiments with speculative decoding in vLLM on AMD Instinct MI300X and MI355X GPUs using the ROCm platform show that output-token throughput gains vary by drafting method, proposal length, model family, and workl…
Engrim has launched a local-first, project-scoped SQLite episodic memory standard that bridges Google Antigravity, Claude Code, and Cursor without cloud lock-in, compressing thousands of interaction tokens into a thin wo…
Runway's GWM Worlds 2, released September 3, 2026, is the first general world model to generate real-time interactive video at 720p/24fps with synchronized 48kHz audio, supporting arbitrary actions and multiplayer via Li…
A new integration checklist from Donely outlines ten best practices for deploying AI employees, emphasizing API-first architecture and multi-instance isolation. The guidance cites a 2021 Vanson Bourne study finding that …
Google's open-source MCP Toolbox for Databases, version 1.9.0, now natively supports ClickHouse, enabling AI agents to perform semantic search by automatically converting text queries into vectors using a Gemini embeddin…
Wafer, an AI company, published a curated reading list for learning GPU performance engineering and production inference, ordered from a single inference request to distributed systems and featuring primary sources such …
Researchers introduced Sarathi-Serve, an LLM inference scheduler that uses chunked-prefills and stall-free scheduling to improve throughput-latency tradeoffs, achieving 2.6x higher serving capacity for Mistral-7B on a si…
NVIDIA CEO Jensen Huang declared on X that 'AGI has arrived' after OpenAI released GPT-6 Astra, trained on over 100,000 NVIDIA Grace Blackwell NVLink72 systems, and announced that 400,000 additional GPUs are coming onlin…
South Korea and the United States have finalized an approximately $22.3 billion investment to build a 6.3 GW gas-fired power plant in Encinal, Texas, aimed at powering AI data centers. The project, expected to begin comm…
A developer explains why smart factories need both cloud AI and edge AI, arguing that edge AI enables near real-time decision-making, data filtering, and resilience to connectivity loss, while cloud AI provides centraliz…
Neo4j Labs released @neo4j-labs/nams-ai-provider, a community provider for the Vercel AI SDK that gives any model persistent, cross-session memory backed by the hosted Neo4j Agent Memory Service (NAMS). The package requi…
A developer has introduced x402, an HTTP-native micropayment specification that enables AI agents to pay for external services using stablecoins directly within the HTTP request/response flow. The protocol repurposes the…
OpenAI announced that its research organization now runs 3.1 agent-workdays of compute effort for every human workday, calling the system an 'automated research intern' that can execute multi-day research tasks with mini…
Kioxia presented a Compute Express Link (CXL) module at a Tokyo briefing that uses its XL-Flash NAND to replace some DRAM in AI workloads, claiming it can double memory capacity at the same cost while improving performan…
A Neo4j semantic layer built with Neocarta cut Text2SQL token costs by up to 81% on a 278-table BigQuery Census ACS catalog and up to 54% on a 264-table Databricks catalog, according to a Neo4j solutions engineer's bench…
OpenAI's inference service returned a 'Codex error: Our servers are currently overloaded. Please try again later' message on September 7, 2026, interrupting a user's ordinary day of using the service. The incident highli…
The OKF Agent Memory repository on GitHub has garnered 395 stars and 20 forks, introducing Git-native persistent memory for AI coding agents. This technology enables agents to maintain state across sessions directly with…