Gemini 3.8 Live Learns to Talk Back
Google's new voice models can reason while they speak, switch languages mid-sentence, and finish tasks in the background while the conversation keeps going.
Arun Kumar
Google Hands Your Smart Home to AI Agents
Google Home MCP lets tools like Claude tap the data and controls behind your smart home, not just flip a light switch.
Baking Safety Into Open AI Models
Anyone can strip the guardrails off an open AI model. Baseten's new research arm wants safety built in from the start, not bolted on.
Plugin4Shell Bug Hit AI Coding Agents
A flaw in Codex, Claude Code, Gemini CLI and Copilot let attackers swap a trusted plugin for malicious code, no developer click required.
DeepSeek's New Flash Model Goes Small
DeepSeek says V4.1-Flash is the smallest model in a new architecture family, with vision built in from the start.
OpenAI's 10,000-Agent Math Claim
OpenAI says a swarm of agents cracked a Navier-Stokes result, but the proof, the details, and the field's verdict are all still missing.
A Map of Every DNA Typo
DeepMind precomputed the molecular effects of all 9 billion possible single-letter DNA changes and is letting researchers browse them for free.
GPT-6 Astra Beats Portal by Itself
OpenAI's new flagship model played through all of Valve's Portal on its own, and one hobbyist caught the whole 24-hour run.
OpenAI's GPT-6 Astra Hits a Cyber Red Line
OpenAI says its new flagship crossed a "Critical" cybersecurity threshold, a first for the company and a puzzle for every enterprise using AI agents.
AI Security Gets a $100M Bet
As companies scramble to keep their AI agents from going rogue, HiddenLayer just raised $100M to guard the machines.