Anthropic makes Claude Opus 5.5 the default and says to drop "think hard"
Opus 5.5 is now the default on Pro, Max, Ultra and Team plans. Anthropic says older prompting habits slow it down, and it recommends starting at Medium effort.
Saturday, November 14, 2026
4 stories, 17 quick hits. Read that day's issue.
Anthropic's Claude Meetings beta lets Claude join Google Meet calls and take notes
Claude can now sit in on a video call, answer questions, write notes and build docs while the meeting runs. The beta is rolling out on Team and Enterprise plans.
llama.cpp and ggml reach version 1.0.0 with a stable API promise
The open-source engine for running AI models on your own hardware now follows strict version rules. A breaking change to its public C API will mean a new major version.
OpenAI launches ChatGPT Wallet so its agents can buy things after you approve
The feature lets OpenAI's dots agents check out with a saved card after you approve with a fingerprint or Face ID check. It is starting with some US ChatGPT Pro subscribers.
Quick hits from this day
- Gemini 4 Carbon tops MathArena but costs 15 times more than GPT-6.1 Sol.The Decoder, New models
- Runway releases Praxis-1, an Apache 2.0 robotics model built on video training.The Decoder, Build
- Wired reports low-skill attackers are using AI tools to drive a ransomware surge.Wired, Rules and safety
- Report alleges Nvidia funded AI startups to buy its GPUs and create scarcity.Hacker News, Chips and data centers
- Amazon plans a World Product Graph indexing products sold anywhere online, Reuters reports.The Verge, Agents
- The Verge tests Claude Duet, Anthropic's consumer agent, and finds it helpful.The Verge, Agents
- ElevenLabs launches ElevenAI, a meeting notetaker and dictation app for teams.TechCrunch, Work and study
- Anthropic's Claude for Manufacturing connects to Fusion, NetSuite and Infor with 17 skills.The Decoder, Agents
- AI helps decode Kurt Gödel's notebooks, reportedly holding hundreds of proofs.Hacker News, Science
- Claude.ai, Claude Code and the Claude platform saw elevated errors.Hacker News, Your AI assistants
- Microsoft 365 recovers from a 24-hour global outage caused by a bad update.Wired, Your AI assistants
- The Atlantic looks at AI book scams on Amazon eroding reader trust.Hacker News, Work and study
- Nvidia CEO Jensen Huang resigned amid a scandal with China security questions.TechCrunch, Chips and data centers
- NPR explores why Google is cutting Search access for AI-free schools.Hacker News, Work and study
- US public-sector unions win AI guardrails as 70% of agencies use generative AI.The Decoder, Rules and safety
- Ai2 releases XYBench to test if models address the real problem behind questions.Hugging Face, Science
- OpenAI disbands its robotics team as general models start running robots.Hacker News, New models