OpenAI replays real chats to test new models
OpenAI’s Deployment Simulation uses 1.3 million consented historical conversations to predict how unreleased models will behave in production, including agentic tool-use failures.
OpenAI’s Deployment Simulation uses 1.3 million consented historical conversations to predict how unreleased models will behave in production, including agentic tool-use failures.
SoftBank and OpenAI are selling an AI security service that scans, ranks and helps patch vulnerabilities for Japan critical infrastructure companies.
RiskIQ veterans emerged from stealth with Ent, a security startup that wants endpoint tools to judge intent before damage is done.
Bland says its voice agents handled 175 million calls last year and now processes 3.5 million calls a week.
Limitless Labs raised $20 million to build a CAM agent that captures machinist expertise and turns CAD files into safer tool paths.
DeepSeek closed its first outside round at a valuation above $50 billion, but commercial backers get no voting power.
AI chip demand has pushed TSMC capacity toward 2028, sending BYD, Google, AMD and Tesla to explore Samsung foundry capacity.
The former Alibaba Qwen lead is building a new lab around world models and embodied AI instead of another chatbot.
Days after its IPO, SpaceX exercised an all-stock option to acquire Anysphere, the company behind Cursor.
Alibaba's Qwen team released three embodied AI models, with RobotManip lifting cross-robot transfer success from 7.5% to 23.9% while RobotNav and RobotWorld cover movement and prediction.