Lobster mode 🦞

Jensen Huang thinks every company should have an OpenClaw strategy

March 18, 2026

Welcome back! Moltbook just updated its rules about what you and your AI can do. The viral social network for AI agents (recently acquired by Meta) now requires agents to verify their identity and stay tethered to human owners. 

It looks like, for now, agents won’t be running wild developing their own secret languages or organizing without their humans in the loop. What do you think? Hit reply and let me know: Was this the right move, or an overstep?

Nvidia Launches NemoClaw, Unveils GTC 2026 Vision

Via TechCrunch / JOSH EDELSON / AFP / Getty Images

Jensen Huang thinks every company should have an OpenClaw strategy. And Nvidia is here to provide it. At his GTC keynote on Monday, Huang announced NemoClaw—an enterprise-grade AI agent platform built on top of OpenClaw, the popular open-source framework for building and running AI agents locally.

What NemoClaw is: The idea is to turn OpenClaw into a secure platform that businesses can tap into with one command, giving them control over how agents behave and handle data—with security and privacy features baked in. Nvidia worked with OpenClaw creator Peter Steinberger to develop it. 

How it works: Users will be able to tap any coding agent or open-source AI model—including Nvidia's NemoTron models—to build and deploy AI agents. The platform is hardware agnostic and integrates with NeMo, Nvidia's AI agent software suite. For now, Nvidia is describing NemoClaw as an early-stage alpha release.

More highlights from GTC: On top of the big NemoClaw announcement, Nvidia revealed plenty of other innovations and partnerships at the event.

  • DLSS 5 launches this fall, using generative AI for real-time neural rendering—Huang called it the biggest graphics leap since real-time ray tracing in 2018.

  • Nvidia introduced Space-1, a Vera Rubin module delivering 25x more AI compute than H100 for orbital data centers.

  • Huang said AI chip demand could hit $1 trillion by 2027, doubling earlier estimates.

  • Plus, much more, which I share in my video below ⤵️

The bigger picture: From agentic enterprise platforms to photorealistic gaming to literal space computing, Nvidia is setting the pace for AI across every layer of the stack. GTC was a declaration that Nvidia intends to be the infrastructure company for the next decade of computing.

The Dictionary Takes OpenAI to Court

Encyclopedia Britannica and Merriam-Webster have filed a lawsuit against OpenAI, alleging "massive copyright infringement." Britannica, which owns Merriam-Webster, says OpenAI scraped nearly 100,000 of the dictionary’s online articles to train LLMs without permission.

What the lawsuit claims: Britannica accuses OpenAI of violating copyright in three ways: 

  1. Using its content as training data

  2. Generating outputs that contain verbatim reproductions of its articles

  3. Using its content in ChatGPT's retrieval augmented generation (RAG) workflow. 

The publisher also alleges OpenAI violates the Lanham Act—a trademark statute—when ChatGPT generates hallucinations and falsely attributes them to Britannica. 

Who else is suing: Britannica joins a growing list of publishers pursuing legal action against OpenAI. Dozens of media outlets—including The New York Times, Ziff Davis (owner of Mashable, CNET, IGN, and PC Mag), the Chicago Tribune, Denver Post, Toronto Star, and Canadian Broadcasting Corporation—have filed similar suits. A separate Britannica lawsuit against Perplexity is still pending.

What’s next: There's still no strong legal precedent establishing whether using copyrighted content to train an LLM is infringement. In one recent case, Anthropic convinced a federal judge that using content as training data was transformative enough to be legal—but the judge still ruled Anthropic violated the law by illegally downloading millions of books rather than paying for them, resulting in a $1.5 billion class action settlement.

Just goes to show that the rules around copyright and AI are still very much being written.

Meta Closes $27 Billion Infrastructure Deal With Nebius

Meta signed a new long-term agreement to spend up to $27 billion on AI infrastructure provided by Dutch company Nebius, an emerging European player in AI cloud computing founded in 2022. Over the next five years, Nebius will provide $12 billion of dedicated capacity across multiple locations and up to $15 billion of additional compute capacity.

A growing force: Nebius also inked a deal to deliver computing resources to Microsoft worth up to $19.4 billion over five years in September. Last week, Nvidia announced it would invest $2 billion in the company.

The bigger picture: Meta said its AI-related capital expenditure will hit between $115 billion and $135 billion this year—part of a combined $700 billion in spending by hyperscalers including Amazon, Alphabet, and Microsoft. The race to build out AI infrastructure continues at full speed, and the companies locking in compute capacity today are betting it will be the scarcest resource of the next decade.

Nebius Token Factory is designed for teams going beyond APIs.

Deploy and scale open-source LLMs, fine-tune them to your use case, and manage access, data, and cost as your product grows. Everything you need to build serious AI, without stitching together tools.

Your always-on business intelligence analyst

Via Pulse AI

Pulse AI is an AI-powered analytics platform that connects to hundreds of data sources to instantly build interactive dashboards and answer questions. 

How you can use it

  • Ask questions in plain language and get real-time answers with follow-up queries

  • Automate detection of opportunities or risks across your data

  • Turn scattered spreadsheets into prioritized, actionable recommendations

  • Keep uploaded data private and excluded from model training

Pricing: Free and paid

The backend platform built for AI coding agents

Via InsForge

InsForge is an AI-native backend platform that gives coding agents everything they need to build, manage, and deploy full-stack apps autonomously. 

How you can use it

  • Let AI coding agents scaffold, operate, and ship backends without manual API key setup

  • Provision resources, inspect schemas, call models, and deploy via MCP agent interface

  • Prototype faster with fewer tokens and errors than traditional BaaS

  • Enable one-click deployments for agentic workflows

Pricing: Free and paid

Jobs, announcements, and big ideas

  • Google rolls out Gemini personalization features to free users.

  • Mamba-3 claims to outperform Transformers with lower latency and modest accuracy gains.

  • Mistral AI debuts Forge, an enterprise platform for building AI models trained on proprietary data.

  • Midjourney opens alpha testing for V8, its fastest image model yet with improved controls.

  • Anthropic launches Dispatch, a research preview that lets Claude run background tasks across apps.

  • Senators urge ByteDance to shut down Seedance 2.0 over intellectual property concerns.

Behind the GTCenes: Sharing what I found interesting from Day 1 of Nvidia GTC.

That’s a wrap! See you Friday for more.

—Matt (FutureTools.io)