AI
Breaking AI announcements, model launches, company news
The lead
Latest first
Browse AI by topic
Open-Source AI
Hugging Face, GitHub, Ollama, local models, agent frameworks
Explore 55 postsModel Launches
Every AI model release that matters: benchmarks, context windows, pricing, and honest first takes.
Explore 43 postsAI Agents & MCP
Autonomous AI agent frameworks, multi-agent systems, and real deployments shipping today.
Explore 29 postsAI Use Cases
Real examples of what people are doing with AI
Explore 22 postsResearch
Peer-reviewed AI research distilled: what was proven, what it changes, and what it does not.
Explore 10 postsPolicy & Safety
AI legislation, safety frameworks, and court cases shaping how AI gets deployed.
Explore 1 postAI Tools
Curated directory of AI tools by category
Explore 1 postLocal AI
Local AI installs and private inference: Ollama, llama.cpp, and self-hosted setups.
ExploreLatest in AI
GitHub lets admins revoke credentials by type
GitHub now lets enterprise and org admins revoke one credential type at a time during a breach, containing damage without locking out…
AWS AgentCore lets AI agents spend on their own
Amazon Bedrock AgentCore payments is generally available, adding MPP support and spending caps so enterprise AI agents can pay for APIs and…
AWS drops NVIDIA’s 30B agent model into SageMaker JumpStart
NVIDIA and AWS made Nemotron 3.5 Lightning, a 30-billion-parameter open model for agent workloads, available on SageMaker JumpStart without self-managed serving.
Axonius isolates each tenant inside Bedrock AI agents
Axonius runs a dedicated Bedrock AgentCore runtime per customer so agents analyze security data without crossing tenant walls. SaaS vendors get the…
Apple: native-language AI reasoning nears English, with risks
Apple's GRPO study shows training models to reason in a native language nearly matches English, but can quietly regress other languages and…
Panchagnula study curbs political bias in LLM summaries
Tejaswi V. Panchagnula and five co-authors show recursive self-correction lifts LLM political-neutrality scores from 2.14 to 4.56 after adversarial bias injection.
Cursor flaw let a harmless command run attacker code
Pillar Security found CVE-2026-22708 in Cursor: shell built-ins bypassed an approval allowlist, so a command a developer approved could execute attacker code.
OpenAI pledges 35,000 jobs for Ohio AI campus
OpenAI secured an 8-gigawatt Ohio data center, promising 35,000 construction jobs, $160M in local benefits, and free Codex credits for students.
Wiola 13M swaps the standard block for spiral attention
Wiola 13M, a 13-million-parameter model, replaces the usual transformer block with spiral attention and a butterfly feed-forward to run efficiently on device.
Ethical RL agents fail where averages hide harm
University of Waterloo researchers show ethical RL agents post clean averages while harming in single episodes; per-episode training and evaluation close the…