AI Infrastructure news and analysis on Web Pulse: 32357 curated articles tracking the latest AI Infrastructure developments, tools, and research, updated continuously from vetted sources.
Snowflake unveiled Horizon Context, a new set of semantic and metadata-management capabilities now in preview, at its annual Snowflake Summit conference on Tuesday. The offering, part of the Horizon Catalog governance su…
Perplexity announced a new hybrid AI feature for its Perplexity Computer platform that will automatically split tasks between local on-device models and cloud-based frontier models. The system will handle sensitive data …
AMD's MI300X accelerator, with 192GB of HBM3 memory and roughly half the list price of NVIDIA's H100, remains underutilized due to software incompatibilities. As of early May 2026, running vLLM with DeepSeek-V4-Flash on …
Microsoft announced Project Solara, a platform for AI agents, at Build 2026, demonstrating it on a smart display and a smart key badge reference device. The platform, built on a fork of Android, is designed to replace tr…
Microsoft released the Surface RTX Spark Dev Box, a compact desktop computer designed for Windows developers, filling the void left by Qualcomm's canceled Snapdragon Dev Kit. The device combines Nvidia's RTX Spark Arm pr…
GitHub released a new Copilot desktop app in technical preview at Microsoft Build, designed as an agent-native control center for developers managing multiple AI agents in parallel. The app provides a unified "My Work" v…
At the ACM CAIS conference in San Jose, a developer presented a paper on simulation-driven resilience for agent-first data systems and observed that much of the conference's work remained exploratory and anecdotal. The d…
A developer wrote a fused decode-attention kernel that ran 2.2× faster than the baseline in microbenchmarks, but when integrated into a HuggingFace `generate` call for an RL training loop, the decode step became nearly 3…
GitHub Copilot now runs inside secure, isolated sandboxes both locally and in the cloud, giving developers and enterprises a controlled environment for agentic workflows. The sandboxes provide restricted access to filesy…
Google Cloud Storage has introduced Model Context Protocol (MCP) servers that enable AI agents to directly access and process unstructured data stored in GCS, turning passive objects into actionable context for reasoning…
Microsoft unveiled Work IQ, a new enterprise platform designed to shift IT operations from human-driven integration to an agent-first model where AI agents dynamically discover and use data structures at runtime. The off…
Microsoft's Applied Sciences Group announced Project Solara, a new software platform and hardware ecosystem designed for agent-first computing experiences. The platform aims to create specialized computers that leverage …
GitHub COO Kyle Daigle detailed how the platform is adapting to a 1,400% surge in code shipped by AI agents in 2026, which is straining infrastructure originally built for human-paced development. Daigle addressed challe…
Crossmint launched a public agentic card payments API using Visa Intelligent Commerce and Basis Theory, enabling eligible US Visa cardholders to authorize AI agent payments with tokenized credentials and spending control…
Microsoft announced the Surface RTX Spark Dev Box at Build today, a desktop device powered by NVIDIA's RTX Spark chip designed for sustained AI workloads like long-running training jobs and local model fine-tuning. The D…
Perplexity introduced Search as Code (SaC), a new reference search architecture that replaces traditional monolithic search pipelines with programmable primitives accessible as SDKs within agent harnesses. The company ar…
Microsoft announced the Microsoft Execution Containers (MXC) SDK, a new cross-platform execution layer for AI agents on Windows and WSL, at Build 2026. The SDK provides policy-based containment to limit what autonomous a…
Oxide engineer Cliff L. Biffle released iddqd, a Rust library for maps where keys are borrowed from values, to solve correctness issues in Oxide's Omicron control plane software. The library uses unsafe Rust code to main…
New York's 26-year-old sales tax exemption for internet data centers, originally estimated to cost the state $9 million annually during the dial-up era, is now allowing AI giants to avoid hundreds of millions in taxes on…
Kapa, a company building AI assistants for technical documentation, developed a method to incorporate images into its RAG pipeline without incurring per-query vision costs. The firm indexes images by describing each one …