OpenAI's ChatGPT Wallet lets its agents buy things after you approve
Today in 5 minutes
AI agents can now pay for things for you, but you still approve every checkout and some big stores like Amazon block them. If you use Claude at work, its default model changed today, along with Anthropic's advice on how to prompt it.
- OpenAI launched ChatGPT Wallet, which lets its agents buy things with a saved card after you approve each checkout.
- Anthropic made Claude Opus 5.5 the default on paid plans and advises dropping prompts like "think hard."
- Anthropic released Claude Meetings in beta, which lets Claude join Google Meet calls and take notes.
- The llama.cpp team released version 1.0.0 of llama.cpp and ggml and promised a stable API.
OpenAI launches ChatGPT Wallet so its agents can buy things after you approve
OpenAI launched ChatGPT Wallet on Friday. It lets its dots agents buy things for you with a saved card. At first it goes to some US ChatGPT Pro subscribers aged 18 and up who have eligible US-issued cards. Before each checkout you see the items, the total and the card, and you approve with Face ID or a fingerprint. A payment token keeps your real card number from both the agent and the store, and OpenAI says card rewards and dispute protections still apply. Some retailers block agents, including Amazon, so OpenAI published a playbook to help stores accept agent purchases.
What it means for you If you are a US ChatGPT Pro subscriber, you may now be able to let a dot agent buy things for you, and you still approve every purchase yourself. If you are on another plan or outside the US, nothing changes yet, and some stores, including Amazon, may block the agent anyway.
Useful forEveryone
Read the storyAlso today
Anthropic makes Claude Opus 5.5 the default and says to drop "think hard"
Anthropic has made Claude Opus 5.5 the default on its Pro, Max, Ultra and Team plans and says old prompting habits are holding it back. The company advises dropping phrases like "think hard," saying the model reasons on its own and started responding faster without them in its tests. It recommends starting at Medium effort, the default, and says Opus 5.5 on Medium matched Opus 5 on High internally. Other tips: put the full goal in one message, attach charts as files, ask for finished documents directly, and name specific design styles to avoid. All performance and cost claims are Anthropic's own.
What it means for you If you are on a Pro, Max, Ultra or Team plan, you are now using Opus 5.5 by default, and Anthropic says you can cut "think hard" phrases and stay on Medium effort unless results fall short. If you often run Sonnet at "Extra High" or "Max" effort, Anthropic says you end up paying about as much as you would with Opus 5.5.
Useful forEveryone
Read the story
Anthropic's Claude Meetings beta lets Claude join Google Meet calls and take notes
Anthropic has launched Claude Meetings in beta for Team and Enterprise plans. Claude can join Google Meet calls as a visible participant, answer spoken or typed questions in writing, use connected tools with permission, and build docs and whiteboards that teammates edit live. It writes notes and action items and saves a transcript, summary and screen descriptions, but no audio. By default it skips calls with outside guests. Organizations using zero data retention, HIPAA-ready setups or customer-managed keys can't use it yet. Notes, transcripts and most answers are free during the beta, and in-call artifacts are free through February 10, 2027.
What it means for you If your team is on a Claude Team or Enterprise plan and meets on Google Meet, Claude Meetings is starting to roll out to you. On Enterprise, an admin has to turn it on. During the beta, Claude's notes and transcripts are free, and docs it makes in calls are free through February 10, 2027. Answers that use connected tools count toward your plan's usage. If you meet on another video platform, or your organization uses zero data retention, HIPAA-ready configuration or customer-managed encryption keys, it is not available to you yet.
Useful forEveryone
Read the story
llama.cpp and ggml reach version 1.0.0 with a stable API promise
The llama.cpp team has released version 1.0.0 of llama.cpp and its tensor library, ggml, according to a Hugging Face blog post. The project lets people run AI models locally on hardware from laptops to servers. With 1.0.0, the team adopts semantic versioning: breaking changes to the public C API mean a major version bump, new features and model support come in minor releases, and bug fixes come in patches. The team says llama.cpp supports more than 150 model architecture variants. It ships through Docker, Debian, Homebrew, conda-forge, MacPorts, Winget and an XCFramework. The team says NVIDIA, AMD, Qualcomm, Arm, IBM and Intel provide hardware support as it adds self-hosted test runners.
What it means for you If you build apps on llama.cpp or ggml, you can now check the version number to tell whether an update might break your code: under the team's rules, only a major version bump should.
Useful forBuilders
Read the storyNew tools to try
Quick hits
- Microsoft 365 recovers from a 24-hour global outage caused by a bad update.Wired
- Claude.ai, Claude Code and the Claude platform saw elevated errors.Hacker News
- ElevenLabs launches ElevenAI, a meeting notetaker and dictation app for teams.TechCrunch
- Anthropic's Claude for Manufacturing connects to Fusion, NetSuite and Infor with 17 skills.The Decoder
- Wired reports low-skill attackers are using AI tools to drive a ransomware surge.Wired
- US public-sector unions win AI guardrails as 70% of agencies use generative AI.The Decoder
- Gemini 4 Carbon tops MathArena but costs 15 times more than GPT-6.1 Sol.The Decoder
- Amazon plans a World Product Graph indexing products sold anywhere online, Reuters reports.The Verge
Did this save you time?
Reply to the email with one thing we should cover, or write to kortexnotebooklm@gmail.com.
AI-assisted and checked against sources. How we work. Want a slot like the ones above? See sponsoring. Paid blocks are labeled Sponsored and never change what we cover.
Get Kortex Daily in your inbox
AI news you can use, in five minutes. Free, every weekday.