The New Stack

131 articles in the Truth Foundry index. Each links out to the original.

• Meta's Iris AI Chip Launch Marks Major AI Infrastructure Shift · Meta announces its first in-house AI chip, Iris, to enhance control over AI infrastructure and reduce costs
• AWS Builds Static Stability into EKS to Survive Gray-Zone Failures · AWS engineered static stability into EKS to prevent single-zone network blips from cascading into regional outages.
• OpenAI Merges Codex into ChatGPT Work to Challenge Claude Cowork · OpenAI integrates Codex into ChatGPT Work, creating a unified agentic platform for knowledge workers.
• OpenAI Launches GPT-5.6 with Sol, Terra, Luna Models Rivaling Anthropic · OpenAI releases GPT-5.6 in three tiers, with Sol matching Anthropic's Fable 5 at half the cost.
• DevRev Launches Open Enterprise AI Benchmark to Test Real-World Data Complexity · DevRev releases an open benchmark measuring enterprise AI agents on data organization and retrieval rather than pure reasoning.
• Nvidia CEO Jensen Huang Declares Traditional Coding Dead in Favor of AI Agents · Nvidia CEO Jensen Huang states engineers are shifting from writing code to building AI agents.
• CISA SBOM Guidance: Five-Minute Sniff Test for Hardened Container Images · CISA mandates deep SBOMs; teams use quick checks to verify container image transparency and security.
• Krumware's Epinio Bridges Local-to-Cluster Kubernetes Gap for Developers · Krumware's Epinio engine standardizes Kubernetes deployments by unifying local and production environments.
• Autonomous AI Pipeline Poisoned Vector Store via Hallucination Loop · A fintech RAG system failed because probabilistic extraction and validation created a hallucination loop in the vector database.
• Elon Musk Unveils Grok 4.5 as Faster, Cheaper Opus-Class Model · SpaceXAI launches Grok 4.5, an Opus-class model optimized for speed and efficiency.
• Meta's AI Chief Claims Parity With OpenAI But Offers No Benchmarks · Meta's Alexandr Wang claims Watermelon model matches GPT-5.5 but provides no verification data.
• Enterprises to Rely on AI Agents for Root Cause Analysis Within Two Years · 85% of organizations use GenAI for observability, with adoption expected to reach 98% in two years.
• Refiant Launches 10-Million-Token Protea Model Using Swarm Optimization · Refiant introduces Protea, a 10-million-token AI model leveraging nature-inspired swarm optimization for efficient long-context processing.
• Entire Launches Distributed Git Network to Handle AI Agent Traffic · Entire unveils a decentralized Git network designed to prevent AI agent fleets from overwhelming centralized servers.
• Coinbase Cuts AI Costs in Half by Deploying 1,200 Agents and Multi-Model Routing · Coinbase slashed AI spend by 50% by routing work across multiple models and automating decisions with 1,200 agents.
• AWS Engineers Demonstrate OpenTelemetry and OpenSearch for Agentic AI Troubleshooting · AWS engineers showcase how OpenTelemetry and OpenSearch unify observability for agentic AI workloads.
• Vercel Acquires Better Auth to Enable AI Agent Identities · Vercel acquires Better Auth to solve AI agent identity and permission management
• Anthropic Extends Fable Model Subscription Access to July 12 · Anthropic delays Fable model paywall, extending free subscription access by five days.
• Anthropic Moves Claude Cowork to Cloud for Always-On Task Automation · Anthropic launches cloud-based Cowork, enabling scheduled tasks to run even when devices are offline.
• Invisible Organizational Data Risks AI Agent Decision Accuracy · AI agents fail when they cannot access undocumented reasoning and institutional judgment hidden in human workflows.
• Replacing Code Review with Intent Verification and AI Slop Registers · Traditional code review is obsolete; replace it with intent verification and automated invariants.
• New Study Debunks Fear That AI Coding Agents Crowd Out Open Source Contributors · Research shows AI coding agents increase code complexity but do not reduce newcomer participation in established open-source projects.
• Microsoft, Google, Cloudflare Move Quantum-Safe Deadline to 2029 · Major tech vendors accelerate post-quantum cryptography transition deadline to 2029 due to advancing quantum threats.
• Token Efficiency, Not Just Cheap Models, Will Save AI Budgets · Agentic AI workflows require token-efficient architectures to control rising operational costs.
• Apple Ships Official MCP Server for Safari to Enable AI Agent Control · Apple integrates a Model Context Protocol server into Safari to allow AI agents direct browser control.
• Chicago Cargo Theft Exposes Physical Supply Chain Risks in AI Boom · Stolen data center equipment highlights emerging physical security vulnerabilities in AI infrastructure supply chains.
• Microsoft Spends $2.5 Billion to Fix AI Strategy Mistake · Microsoft launches $2.5 billion Frontier unit to help enterprises swap AI models instead of locking into one.
• Godot Engine Bans AI-Generated Code to Protect Mentorship Model · Godot Foundation prohibits AI-generated code to preserve the educational value of pull request reviews for human contributors.
• OpenClaw App Launches as Remote Control for Off-Device AI Agents · OpenClaw's new mobile app connects to AI agents running on remote servers instead of local phones.
• CI/CD Pipeline Flaw 'Cordyceps' Exposes Open-Source Supply Chains to Hijacking · A CI/CD configuration flaw dubbed 'Cordyceps' allows unauthenticated users to hijack trusted workflows across major tech organizations.