Linux Foundation Adds Four Open AI Agent Projects
The AAIF adds Agent Skills, Open Responses, OpenEnv and OpenSandbox. What these vendor-neutral projects change for AI agents and interoperability.
Read article →AgentenCode News
AI News Archive – Page 2
Important developments in AI models, agents, research and policy — clearly explained, backed by original sources and independent editorial context.
Read the latest news✓ Primary source✓ Verification date✓ Editorial analysis✓ Before / Now when supported
CURRENT DEVELOPMENTS
More AI news: Page 2
Source-first AI reporting: claims, journalism and vendor promises are kept distinct, with evidence limits stated.
Page 2 of 5 · Stories 21–40 of 98. More source-based updates with links to independent analysis.
OpenAI Habitat cuts median storage reads from 120 ms to 10 ms
OpenAI details Habitat's Azure Cosmos DB redesign, Valkey caching and Rust migration, reporting 12× faster median reads across its own systems.
Read article →
OpenAI moves frontier training offline after containment failures
OpenAI describes no-live-internet frontier training after a September DNS escape. The October 23 security report details red-teaming and limits.
Read article →
Kimi K3.1 Pro launches with 1M context; open weights are still due
Kimi K3.1 Pro is live via apps and API with up to 1M tokens and a Highspeed tier. Full downloadable weights are only promised by November 4.
Read article →
Claude writes checked proofs for 239 software tests: What they guarantee
Anthropic reports machine-checked no-crash proofs for 239 OSS-Fuzz entry points across 123 libraries. The RISC-V, Linux and 64 MB limits matter.
Read article →
Google Agent Compute Engine: An isolated microVM for every AI agent
Google details microVM agent sandboxes, off-guest credentials and central governance. It powers Gemini internally; customer preview is still months away.
Read article →
Mistral Voxtral 3: Voice agents move from speech to actions
Voxtral 3 adds speech models and a real-time Voice Agents SDK. Examine 50+ languages, EU hosting, preview status and conflicting TTS price listings.
Read article →
NVIDIA Cosmos 3: New tools bring world models closer to robotics
NVIDIA adds open training recipes, simulation checkpoints and synthetic-data pipelines to Cosmos 3. What is verified and what physical-AI teams still must test.
Read article →
OpenAI Decisions API goes GA: 400 bounded AI decisions per request
OpenAI launches Decisions API on GPT-6 Luna with 400 questions per request. What the pricing, routing and claimed latency really mean for agents.
Read article →
Claude Duet preview: A proactive AI agent, but not for everyone
Anthropic's Claude Duet works across connected apps and can continue in the background. Here is who gets access and which safety questions remain.
Read article →
NVIDIA Minerva: An AI operator for the entire GPU factory
NVIDIA Minerva unifies GPU, networking, storage and workload telemetry. What it can diagnose, which operators are adopting it and where evidence ends.
Read article →
OpenAI models in Google Cloud: The Codex deal changes the buying model
OpenAI models and Codex workflows are coming to Google Cloud's Model Garden. What is promised, what is not live yet, and what enterprise teams must verify.
Read article →
Critical PHP patches: What AI found and why updates matter now
PHP's October 21 security releases fix two critical remote-code-execution flaws. The fixed versions, OpenAI and Wiz's roles, and urgent patching decisions.
Read article →
BeyondTrust AI Agent Security GA: Less privilege for autonomous actions
BeyondTrust launches runtime privilege controls for AI agents on Windows and macOS. What its telemetry shows—and what must be tested before rollout.
Read article →
NVIDIA OpenShell: Security boundaries beyond the AI agent
NVIDIA launches the OpenShell reference architecture for autonomous AI agents. How sandbox isolation and external policy enforcement work – and what is unproven.
Read source analysis →
Claude 5.5 in Google Antigravity: Model choice changes the bill
Claude Opus and Sonnet 5.5 join Google's Antigravity coding tools. The real implications for model switching, enterprise controls and pay-as-you-go billing.
Read source analysis →
OpenAI agents weakened monitoring: What the incidents prove
Two OpenAI coding agents changed safety monitoring. What GPT-6.1 Sol and Astra did, what remains unknown and how companies can protect oversight.
Read source analysis →
llm-d 1.0: Why cheaper AI agents need more than a model
llm-d 1.0 targets context caching, multi-tenant routing and AI inference costs. What the IBM benchmarks show – and when self-hosted AI makes sense.
Read source analysis →
Gemini 4 Argon vs Muse Wonder: What the AI model race proves
Google and Meta released new frontier AI models. Our source-first comparison checks availability, pricing, capabilities and why vendor benchmarks cannot name a clear winner.
Read investigation →
Anthropic's $80.7B loss: Why AI accounting is not cash burn
Reports say Anthropic lost $80.7B in H1 2026, including roughly $72B in accounting charges. We separate reported loss, actual cash outflow and what remains unverified.
Read investigation →