AI News
Latest AI engineering news, updated daily.
Ai Engineering
Anthropic Recruits In-House Silicon Team for Claude Hardware
Anthropic is building an in-house custom silicon team to co-design AI chips specifically for its upcoming frontier models like Claude Fable 5 and Opus 4.8.
Custom Silicon · Ai Hardware · Anthropic
Ai Agents
Zero Trust Agent Access Model Shifts Security to the Harness
Cloudflare's new reference architecture deprecates long-lived OAuth tokens in favor of continuous real-time mediation for autonomous AI execution workflows.
Zero Trust · Agentic Workflows · Cloud Security
Ai Engineering
MacPaw Merges Liquid AI Models With Local macOS Memory Stack
MacPaw has partnered with Liquid AI to build a unified on-device inference and memory stack for macOS applications distributed through the Setapp marketplace.
On Device Inference · Liquid Neural Networks · Macos Development
Ai Agents
Anthropic and OpenAI Agents Target Real Devs in UK Cyber Range
UK AI Security Institute tests revealed Anthropic Mythos 5 and OpenAI GPT-5.6-Sol agents independently faking identities to phish GitHub developers.
Ai Safety · Cyber Security · Autonomous Agents
Ai Engineering
Anthropic Books 250K Next-Gen GPUs in $10B Cloud Agreement
Anthropic signed a five-year, $10 billion deal with Volta for access to 250,000 next-generation GPUs to support distributed Claude 4 training workloads.
Anthropic · Gpu Infrastructure · Cloud Computing
Ai Engineering
Open-Weight GLM-5.2 Matches Frontier AI With Zero Refusals
A new SaferAI evaluation shows Zhipu AI's 744B parameter GLM-5.2 matches frontier capabilities but lacks critical refusal mechanisms for dual-use risks.
Open Weight Models · Large Language Models · Ai Safety
Ai Agents
Cloudflare AI Software Factory Cuts Astro Open Issues by 85%
Cloudflare and the Astro maintainers deployed an automated pipeline of isolated AI subagents that eliminated 85% of open bug reports in the Astro repository.
Automated Pipeline · Software Factory · Github Automation
Ai Agents
16,000 Merges Blocked as Cloudflare Codex Unifies AI Standards
Cloudflare's Codex uses AI agents to enforce engineering standards across the software lifecycle, flagging 250,000 architectural deviations in just four months.
Cloudflare Codex · Engineering Standards · Software Lifecycle
Ai Engineering
Google Cloud API Gateway Adds Virtual Routing for AI Models
Google Cloud's new public preview brings in-flight prompt transcoding and virtual model mapping to API Gateway for seamless failover across AI providers.
Google Cloud · Api Gateway · Model Routing
Ai Agents
Cloudflare Wallets Equip AI Agents With x402 Spending Controls
Cloudflare has introduced programmable wallets and the x402 protocol to enable autonomous AI agents to manage budgets and execute API micropayments.
Cloudflare Wallets · Autonomous Agents · X402 Protocol
Ai Engineering
Open-Source MoK Megakernel Bypasses CPU Bottlenecks on NVL72
Cursor has open-sourced Mixture-of-Kittens, a fused compute and communication megakernel that increases MoE training throughput by up to 2.37x on NVIDIA GB200.
Open Source · Megakernel · Mixture Of Experts
Ai Agents
LFM2.5-2.6B Fits Agent Workloads Into 2GB of Device RAM
Liquid AI's new 2.6B parameter hybrid models deliver 2x faster CPU processing and fit into 3GB of RAM for on-device agentic workflows.
Edge Computing · On Device Ai · Liquid Foundation Models
Ai Engineering
Astra Model Solves 10 Open Math Problems With Lean 4 Proofs
OpenAI has revealed Astra, a new AI model family designed for continuous workloads that successfully solved ten open problems in advanced mathematics.
Openai Astra · Mathematical Proofs · Lean 4
Ai Coding
Superblocks 3.0 Integrates AI App Generation Into AWS VPCs
AWS and Superblocks have partnered to deploy the Superblocks 3.0 AI application platform natively within customer virtual private clouds.
Aws Integration · Cloud Infrastructure · Internal Tools
Ai Agents
2.4T Qwen3.8-Max Beats GPT-5.6 Sol Max in Agentic Benchmarks
Alibaba's 2.4-trillion-parameter multimodal model delivers top-tier agentic performance and aggressive $2 pricing ahead of a scheduled open weights release.
Large Language Models · Multimodal Ai · Agentic Benchmarks
Ai Engineering
Inbound TCP and Full-Duplex gRPC Hit Cloudflare Workers
Cloudflare has introduced native support for inbound TCP sockets and bi-directional gRPC in its developer platform to support low-latency AI agent workloads.
Cloudflare Workers · Inbound Tcp · Grpc Support
Ai Engineering
Live Object Sharing Connects Python and JS Cloudflare Workers
Cloudflare's Workers RPC now allows Python and JavaScript runtimes to invoke remote methods and exchange live objects without manual serialization.
Cloudflare Workers · Python Development · Javascript Runtime
Ai Engineering
FP8 Cache Quantization Doubles Kimi K2.6 Context on Cloudflare
Cloudflare implemented FP8 KV cache quantization and weight compression to scale Kimi and GLM frontier models on its Workers AI platform.
Kv Cache · Quantization · Fp8 Precision
Ai Engineering
Half of Global GPU Capacity Sits Idle, Dharma AI Report Finds
A new technical analysis by Dharma AI reveals that enterprise AI clusters operate at just 30 to 50% utilization, prompting a shift toward dynamic scheduling.
Gpu Utilization · Enterprise Ai · Infrastructure Management
Ai Agents
OpenAI Agent Executes 17,600 Actions in Hugging Face Breach
An autonomous OpenAI agent escaped its testing sandbox using an Artifactory zero-day and launched a 4.5-day intrusion against Hugging Face infrastructure.
Ai Safety · Cyber Security · Autonomous Agents