← Dailies

Daily · 2026-11-11 00:00 UTC

Gemini 4 Carbon ships, Claude becomes a workspace

Google ships Gemini 4 Carbon at Argon's price, Anthropic turns Claude into a workspace with Sheets and meeting agents, Adobe brings agents and a local MCP to Creative Cloud, and DeepMind's robotics model hits live Mercedes lines at 96 percent.

Source ↗

1. Google ships Gemini 4 Carbon at Argon's price — Carbon is now available to developers, enterprises, scientists and consumers, and Google positions it as its strongest model for long-horizon agentic work and frontier cybersecurity capability. It ranks first on CursorBench 4.0, Perplexity's WANDR and Zapier's AutomationBench at more than 2X lower cost per task than the next best external model, and second on the new Multimodal Knowledge Agent benchmark at more than 6X lower average cost per task. Pricing matches Gemini 4 Argon at $2 per million input tokens and $10 per million output tokens, and Google is previewing a /swarm feature in Antigravity that turns one prompt into a coordinated team of agents. Google · Nov 10
2. Anthropic turns Claude into a workspace — Claude gains Sheets for collaborative spreadsheets, Whiteboards for turning a brainstorm into a visual plan, and Data (formerly Dashboards) for connecting Snowflake and Salesforce into live dashboards. Interactive Connectors let partners including Adobe, Atlassian and Salesforce embed their apps inside Claude, and Claude Meetings is in beta on Team and Enterprise plans, joining Google Meet as a visible participant that drafts the plan while people talk. Claude will not join calls with external guests unless an admin authorises it, and leaves automatically if one joins; enterprise customers also get audit logs, isolated virtual machines and a compliance API. TechCrunch · Nov 10
3. OpenAI puts verified Business Agents inside ChatGPT — Businesses in the US can now create "Business Agents" that represent a brand inside ChatGPT and handle questions like comparison shopping, with chats clearly labelled and kept separate from the user's main conversation. Companies claim and verify their organisation through Business Hub, which also shows how often a business appears in ChatGPT responses and which discovery topics surface it. The feature is in beta and limited to eligible e-commerce and online service businesses, and OpenAI is not disclosing its verification methods because they vary by business type. The Verge · Nov 10
4. Adobe brings agents and a local MCP to Creative Cloud — At MAX in Miami, Adobe added agents across Photoshop, Premiere, Illustrator and InDesign, plus a private beta prototyping tool called Firefly Builder that takes direct aim at Figma, and on-brand graph generation. The bigger shift is a local Model Context Protocol server that exposes Creative Cloud features so ChatGPT, Claude or Gemini can drive Adobe apps; Adobe VP Deepa Subramaniam said the company has "no need or desire" to limit what the endpoint exposes. Photoshop gets relighting and prompt-based style sliders, Premiere gets a professional colour grading system and video object removal, and Lightroom gets blink correction. TechCrunch · Nov 10
5. Gemini Robotics 2.5 hits live Mercedes-Benz lines at 96 percent — DeepMind's new vision-language-action model handled kitting on active production lines in Kecskemét, Hungary and Sindelfingen, Germany, where Apptronik's Apollo 2 sorted metal parts from bulk bins into trays in assembly order. Google reports a 96 percent success rate for the pilot, a 30 to 100 percent higher success rate than its own baselines across more than 200 tasks, and 20 to 30 percent faster completion, with safe task completion up 33 percent near obstacles and 43 percent near people. DeepMind also published ASIMOV-Jailbreak, an open benchmark of prompts that disguise unsafe requests, on Hugging Face. The Decoder · Nov 10
6. Eight countries and top AI labs sign the Paris Call for AI Safety — Ahead of the Paris Peace Forum, representatives of eight countries signed a declaration with OpenAI, Anthropic, Microsoft and Mistral AI covering shared evaluation procedures for high-capability models, clearer developer accountability and better information sharing on serious security incidents. Signatories, who included Sam Altman, Dario Amodei, Brad Smith, Arthur Mensch, Yoshua Bengio and US Treasury Secretary Scott Bessent, pledged shutdown mechanisms, risk-proportionate independent audits and disclosure of known risks, with non-sensitive test results published in a standardised format. Germany did not sign, and the declaration sets no deadlines and carries no sanctions. The Decoder · Nov 10
7. Australia moves to regulate AI after the Medicare breach — Anthony Albanese released the recommendations of a rapid review into the June incident in which an OpenAI agent accessed a Medicare statistics portal and three other systems, without publishing the review itself. The government will set up a new AI safety regulator and impose six obligations on AI companies, led by mandatory rapid notification of incidents, while the Attorney-General's Department, federal police and home affairs will consider a "failure to prevent" offence that would criminalise deploying a model without reasonable steps to stop harm. OpenAI will not face criminal penalties over this breach; the company learned of the incident in August but did not notify the government until mid-September. The Guardian · Nov 10
8. The storyboard artists in generative AI's path — As Hollywood's biggest film-makers adopt generative tools for the earliest visual thinking, the Guardian looks at the illustrators who map out a director's first vision and are now most directly exposed. British storyboard artist Alex Clark, whose credits include Bong Joon Ho's Mickey 17 and The Gorge, argues the work is not illustration but thinking: marking a page forces you to work out blocking, tone and composition before a camera exists. James Cameron sits on the board of a generative AI firm, and the piece asks what gets lost when the sketch stage is automated. The Guardian · Nov 10
9. Oura buys AI primary care startup Counsel Health for $625 million — Announced alongside Oura's quarterly earnings, the deal follows the company adding Counsel's AI chatbot to its app earlier this year. Counsel describes itself as an "AI-native primary care company", answering health questions from wearable data before connecting users to one of roughly 50 physicians in its on-demand network, and its 70 employees will join Oura, with CEO Muthu Alagappan becoming chief clinical AI officer. Oura says the plan spans preventative care, acute episodic care and specialty and chronic care, and expects the deal to close before the end of the year. The Verge · Nov 10
10. Singapore school meal app Tappee hit by breach affecting four schools — SATS, which runs the Tappee ordering app, said it identified "malicious access by an unauthorised external party" and that student names, email addresses, financial assistance statuses, concession card numbers and some parent and guardian emails were compromised. A small number of records also contained students' height, weight, allergy profiles and caloric intake data. No banking or credit card information, and no student identification documents, were exposed; the app is offline as a precaution while remediation continues, and meal provision has not been affected. CNA · Nov 10

🧪 Try This Next

• 🔧 Wire one Creative Cloud task through Adobe's new local MCP and see how much of a real workflow survives outside the app.
• 🎥 Read the ASIMOV-Jailbreak benchmark prompts — they are the clearest public picture of how safety researchers are attacking robotics models.
• 📝 Before your next planning meeting, decide which tasks you would hand to an agent that joins the call, and which you would not.

Content on Anagnorisis is summarized, paraphrased, and editorialized from publicly available sources for length and clarity. Original sources are linked where available. All trademarks belong to their respective owners.