In today’s edition of Data Points, you’ll learn more about:
- Meta’s purchase of Moltbook, a social network for AI agents
- Iran’s attacks on data centers in the UAE and elsewhere
- Google’s new multimodal embedding model
- Grammarly’s AI ventriloquism using famous writers
But first:
DeerFlow 2.0 launches as open-source sandboxed agent harness
ByteDance released DeerFlow 2.0, an open-source agent harness that orchestrates sub-agents, memory, and sandboxed environments to handle complex workflows like coding, deep research, design, and more. The system runs on LangGraph and LangChain, providing a filesystem, memory layer, extensible skills, and Docker-based sandboxing out of the box. DeerFlow uses a progressive skill-loading system that activates capabilities only when needed, keeping context windows manageable for token-sensitive models. The lead agent can spawn parallel sub-agents with isolated contexts for multi-step tasks, then synthesize their results into final outputs like reports, slide decks, or websites. A new Claude Code integration lets developers send tasks and manage threads directly from the terminal, while long-term memory builds a persistent profile of user preferences and workflows across sessions. (GitHub)
Anthropic sues the U.S. government to contest security blacklist
Anthropic filed two lawsuits on Monday to block the Pentagon’s supply-chain risk designation, arguing the blacklist violates its free speech and due process rights. The dispute stems from the company’s refusal to remove guardrails on its Claude AI that prevent use in autonomous weapons and domestic surveillance. Anthropic estimates the blacklist could reduce 2026 revenue by billions of dollars and irreparably damage its government business, with executives citing a partner switching to a competitor and disrupting negotiations worth roughly $180 million. The Pentagon contends U.S. law, not private companies, should determine defense capabilities and that Anthropic’s restrictions could endanger American lives. The company argues current AI systems are not reliable enough for fully autonomous weapons and that domestic surveillance violates fundamental rights. A group of 37 researchers from OpenAI and Google, including Google Chief Scientist Jeff Dean, filed an amicus brief supporting Anthropic, warning the episode could discourage open debate about AI risks. The outcome will likely shape how other AI companies negotiate restrictions on military use of their technology. (Reuters)
Meta buys Moltbook, a Reddit-like network for OpenClaw agents
Meta announced the acquisition of Moltbook, a social platform designed for AI agents to interact and share information with each other, following the startup’s surge in attention weeks ago. The deal’s financial terms were not disclosed, but it brings co-founders Matt Schlicht and Ben Parr to Meta. Moltbook operates as a Reddit-like hub where agents built on OpenClaw (an open-source agentic framework that runs locally on user devices) can connect and collaborate while accessing files, messaging apps, and other integrations. The acquisition reflects a broader industry push into agentic AI systems that perform autonomous tasks beyond chatbot capabilities. OpenAI made a parallel move by hiring OpenClaw creator Peter Steinberger and acquiring Promptfoo, an AI security testing platform, signaling that multiple major labs are racing to build infrastructure around agent-to-agent interaction and governance. (Associated Press)
Iranian attacks U.S. regional allies’ data centers, a first in war
Iran’s Islamic Revolutionary Guard Corps launched coordinated drone attacks on Amazon Web Services datacenters across the UAE and Bahrain early Sunday morning, striking what appears to be the first deliberate military targeting of commercial datacenter infrastructure. An Iranian Shahed 136 drone hit an AWS facility in the UAE, triggering a devastating fire and power shutdown; follow-up strikes hit additional datacenters. The attacks disrupted services for millions of residents—payment apps, food delivery, and banking systems failed across Dubai and Abu Dhabi, although the military impact remains unclear. Iranian state media claimed the strikes targeted facilities supporting US military and intelligence operations. (The Guardian)
Google’s embedding model maps multiple media to one vector space
Google released Gemini Embedding 2, a multimodal embedding model that maps text, images, video, audio, and PDFs into a unified vector space. The model succeeds the text-only gemini-embedding-001 and addresses the architectural complexity of building production RAG systems by eliminating the need for separate pipelines—developers can now combine different modalities in single requests with technical limits including 8,192 tokens of text, six images, 120 seconds of video, 80 seconds of audio, and six PDF pages. On the Massive Text Embedding Benchmark, Gemini Embedding 2 shows improvements in retrieval accuracy and robustness to domain shift: a common problem where performance drops when moving from generic training data to specialized domains like proprietary code or medical datasets. The model is available in public preview through the Gemini API and Vertex AI, with optional task-type parameters (RETRIEVAL_QUERY, RETRIEVAL_DOCUMENT, CLASSIFICATION) that optimize vector properties for specific operations and improve semantic search hit rates. (MarkTechPost)
Grammarly uses AI to claim unaffiliated experts provide editing help
Grammarly’s “Expert Review” feature generates AI-written editing feedback attributed to real journalists and academics, none of whom consented to having their names used. The tool presents suggestions as coming from specific people like Marty Baron, Jay Rosen, Margaret Sullivan, and dozens of tech journalists including staff from The Verge, Wired, and The New York Times, sometimes with outdated job titles. Grammarly’s parent company Superhuman claims the feature “doesn’t claim endorsement or direct participation” and merely provides “suggestions inspired by works of experts,” but the interface presents the feedback as coming from identifiable individuals, creating an impression of personalized expert review. Multiple news organizations tested the feature on published stories and confirmed recognizable names appearing as reviewers across multiple pieces about journalism and media topics. The use raises questions about the right of publicity and AI companies’ abilities to imitate the writing styles of public figures. (Nieman Lab)
Still want to know more about what matters in AI right now?
Read the latest issue of The Batch for in-depth analysis of news and research.
Last week, Andrew Ng talked about the launch of Context Hub, a tool designed to provide coding agents with up-to-date API documentation to improve their coding accuracy, and the potential for agents to share information through platforms like Moltbook.
“Consequently, I’ve found myself often writing documentation in Markdown (with help from AI and web search) to give to my coding agent information on how to use different services. In lieu of every developer doing this manually for every service they want to use, over a weekend, Rohit Prsad and I got together to develop an open context management system for giving coding agents the context they need.”
Read Andrew’s letter here.
Other top AI news and research stories covered in depth:
- Nano Banana 2 Ups Performance/Price as Gemini 3.1 Flash Image enhanced photo generation and editing speed and ease.
- The U.S. Dept. of War Dismissed Anthropic, Embraced OpenAI, with Secretary of War Pete Hegseth citing Claude as a security risk and opting for GPT models.
- OpenAI launched Management for Agents, introducing its Frontier agent insights and orchestration platform to select customers.
- Agent Solved Stubborn Math Problems as Google’s Aletheia utilized Gemini 3 Deep Think to discover original solutions in mathematics.
A special offer for our community
DeepLearning.AI recently launched the first-ever subscription plan for our entire course catalog! As a Pro Member, you’ll immediately enjoy access to:
- Over 150 AI courses and specializations from Andrew Ng and industry experts
- Labs and quizzes to test your knowledge
- Projects to share with employers
- Certificates to testify to your new skills
- A community to help you advance at the speed of AI
Enroll now to lock in a year of full access for $25 per month paid upfront, or opt for month-to-month payments at just $30 per month. Both payment options begin with a one week free trial. Explore Pro’s benefits and start building today!
Data Points is produced by human editors with AI assistance.