
Anthropic’s AI Now Improves Its Own Alignment Training
Anthropic published a paper showing an AI system can improve its own alignment training, closing safety gaps that took human researchers far longer.
Search the archive
Category archive
Browse the latest posts in this topic and its related archive.

Anthropic published a paper showing an AI system can improve its own alignment training, closing safety gaps that took human researchers far longer.

Hugging Face's Pollen Robotics opened pre-orders for Microduck, a $399 open source robot duck with a full reinforcement-learning stack you can fork and retrain.

OpenAI, Anthropic, and Google just published a joint open letter calling for a surge in AI cyber defense before AI-enabled attacks get out of…

Plaud's new Plaud One earbuds put an AI agent in your ears, with a 4G case that works without your phone. Priced at $249.99,…

Arduino and Qualcomm opened preorders for the Ventuno Q, a $299 edge AI board that runs LLMs and agentic AI fully offline on a…

OpenAI and Broadcom unveiled Jalapeño, a custom AI inference chip that beats Nvidia Blackwell on speed and efficiency at 700W vs 1,200-1,400W on the…

ARIA confirmed on August 25, 2026 that wholly AI-generated tracks will not be eligible for the ARIA Charts starting with the chart dated August…

Most of us running AI coding agents still live in five windows. A real AI development workspace should fix that: issue in the browser,…

NVIDIA has released Nemotron 3.5 Lightning, a 30-billion-parameter open model for specialized agent tasks, alongside an open routing library called NeMo Switchyard.

Netlify's AI Gateway is now generally available on credit-based plans, putting model access, billing, and basic guardrails inside the platform where developers deploy their…

Nvidia's new agent harness research shows the scaffolding around a model now drives accuracy and cost, with big ARC-AGI-3, SWE-bench, and CyberGym gains. Here's…

Google folded Google Antigravity into eligible Gemini Enterprise subscriptions, adding IDE extensions, pooled quotas, budget caps, and audit controls. Here's what ships, what's still…