This website uses cookies
Read our Privacy policy and Terms of use for more information.
Gen Digital's engineers published the line most AI vendors leave off the rate card: reviewing AI output took 20 to 30 percent of their time. Plus Gemini survey analysis in Google Forms, Tencent's Hy4 against DeepSeek on a test Tencent ran itself, and a voice agent for your most scripted queue.
Instinct's $10 billion valuation meets DoorDash's 993 million quarterly orders. Plus shopping agents that can raise the price you pay, Anthropic's $567 billion compute commitment, and a Google model that goes to cyber defenders first.
Google turned on Gemini for BigQuery data in Sheets with no admin control, Microsoft's legal skills arrive enabled in Word, and the fix for 2 PHP flaws rated 9.2 waits on whoever runs your server.
On December 1, Gemini 4 Argon doubles to the exact rate Anthropic sells as 40% cheaper for Claude Opus 5.5, while Nvidia's newest open model runs agents on hardware you already own at $0 per token.
Google's Gemini agent reaches into Gmail, Slack and Microsoft 365 from a single prompt box. Atlassian tied Jira's AI to OpenAI's models. ChatGPT and TikTok moved the sale inside the conversation. LibreOffice shipped with 0 generative AI features on purpose, and watching an agent got 29x cheaper.
OpenAI held back GPT-6.1 Astra for acting without asking, then shipped always-on Dots the next day. Apple, Nvidia, and a $2.94 agent monitor all moved on the lock in the same week.
GPT-6 Sol halves the price of frontier agent work, Google narrates 78 seconds of audio for 2.74 cents, and an AI video startup claims a $1 billion revenue pace. Meanwhile 56% of social users see AI slop often, and half of Gen Z has already blocked a brand over it.
Claude, Muse, and Gemini all got deeper into your accounts this week. IBM, Microsoft's own court filings, Deloitte, and a near-miss military report all measured what happens when nobody checks the output.
DeepSeek prices cache-hit tokens at $0.006 per million, OpenAI stops selling $200 Pro seats with no reopen date, and a community benchmark finds the same model costs 17x more per correct answer depending on what you wrap it in.
