🍭 China’s Claw Crackdown 🦞

Good morning. The intern tried explaining “agentic workflows” to his mom last night.
She asked if that’s the same thing as astrology.

Let’s dive in 👇

🍭 What’s Cookin’:

• Nvidia kicks off GTC 2026 with agentic AI everywhere

• China’s OpenClaw hype hits a security wall

• Atlassian cuts 1,600 jobs to pivot to AI

China
🦞 Raised the Lobster and It's Biting Back

The Bite:

OpenClaw has become a cultural phenomenon in China.

Hundreds of people are lining up outside Tencent and Baidu offices to get it installed on their laptops.

Local governments in Shenzhen and Wuxi are handing out grants worth millions of yuan for OpenClaw-powered startups.

Tencent, Alibaba, Baidu, ByteDance, Moonshot, and Minimax have all released their own "Claw" spinoffs.

Beijing had been actively promoting an AI-positive agenda to grow its economy.

But now, Chinese government agencies and state-owned enterprises are warning employees not to install OpenClaw on work devices.

The stated reasons: data leaks, accidental deletions, and misuse of sensitive information.

The lobster hats are still on. The panic is also setting in.

Snacks:

• ~1,000 people lined up outside Tencent's Shenzhen HQ on a single Friday afternoon — students, retirees, office workers

• Shenzhen's Longgang district offered grants up to 10M yuan ($1.4M) for one-person OpenClaw startups

• ByteDance, Alibaba, Tencent, Moonshot, and Minimax all launched their own "Claw" spinoffs within weeks

• Meta's director of safety and alignment watched OpenClaw delete her inbox after she told it to "confirm before action"

• Agents have been exploited via prompt injection to exfiltrate crypto wallet keys and financial data

Why it Bites:

There's something poetic about this.

Beijing built a national AI strategy on enthusiasm and subsidies.
Officials cheered open-source AI. People showed up in lobster hats.

And then the AI started deleting things.

The irony isn't that OpenClaw is dangerous, because everyone knew that.

Meta had already banned it on work devices.
Security researchers flagged prompt injection risks publicly.
Didn't matter.

China institutionalized the hype before the guardrails existed.

This is what happens when competitive anxiety meets an ungovernable technology.

The same agencies signing OpenClaw startup contracts last month are now quietly telling staff to uninstall it.

The broader signal: agents aren't chatbots.

Chatbots answer. Agents act.

That's a different threat surface entirely.
China found out in real time, in public, with photographers present.

The West is about three months behind the same lesson.
Nvidia recently announces NemoClaw (we’ll cover that this week).
Every SaaS vendor is rushing to go "agentic."

The lobster is coming for your inbox too.

The question is whether the guardrails exist before it gets there.

Astronomer
🧠 Orchestrate LLMs & Agents with Airflow

Anyone can write a wrapper around an LLM in an afternoon.

Shipping AI applications that actually make it to production (and stay there) is a different problem entirely.

The answer is orchestration.

Download the complete guide to GenAI orchestration with practical guidance and code examples.

• Explore patterns like event-driven execution and multi-agent workflows.
• Discover new Airflow 3 features from human-in-the-loop operators to the AI SDK.
• Best practices to keep in mind when building your pipelines.

Steal This Prompt
🎭 3D Funny Caricature

Turn any portrait into a glossy, exaggerated 3D caricature with oversized features, big expression energy, and “this should not work but absolutely does” charm.

Use it to:

• Turn selfies into cursed-but-cute profile pics

• Make creators, founders, or friends look Pixar-adjacent

• Cook up thumbnails, reaction images, or social posts that instantly grab eyeballs

Workflow:

1. Click this link (Prompt)

2. Paste into your image model

3. Replace the #s with your subject or upload your photo

4. Watch it cook into a big-headed little menace

   Try the prompt on Snack Prompt

_ToolBox™
🧰 5 BRAND NEW AI LAUNCHES

🤖 MuleRun
Raise a personal AI agent that learns how you work, then automates research, reports, and tasks while you’re offline.

📸 Glam AI
Turn a photo into viral-ready social content using AI trends, templates, and image/video effects.

🛝 Z.ai
A playground for testing high-performance GLM models for reasoning, coding, and complex agent workflows.

🦊 Masko Code
A desktop mascot that watches your Claude Code agents, surfaces permission prompts, and lets you control sessions without constant tab-switching.

👥 Wendi AI
An AI operating system designed for managers. It helps coordinate teams, tasks, and decisions from one intelligent workspace.

_{Everything Else}
🧠 You Need to Know

🧠 Nvidia Kicks Off GTC 2026
→ Nvidia opened its flagship GTC conference in San Jose, featuring Jensen Huang’s keynote and over 1,000 sessions on agentic AI, robotics, and next-gen computing.

🌟 OpenClaw Becomes GitHub’s Most-Starred Project
→ What started as a solo side project in Nov. 2025 hit the top in ~60 days; OpenAI has since acquired its creator and a new foundation will steward the project.

💼 Atlassian Cuts 1,600 Jobs Amid AI Pivot
→ The software giant is laying off about 10% of its workforce and replacing its CTO as it restructures the company around AI and enterprise growth.

🦞 China Gets Nervous About OpenClaw AI Agents
→ OpenClaw autonomous agents are exploding in popularity, letting users run powerful AI tasks locally—but Chinese authorities are warning about the security risks.

⚖️ Anthropic’s Pentagon Lawsuit Gets Support From AI Workers
→ Nearly 40 employees from OpenAI and Google filed an amicus brief backing Anthropic’s lawsuit against the Pentagon over restrictions tied to military AI use.