Welcome, humans.

Okay, so apparently data centers have become one of the few things Americans are willing to get arrested over.

Futurism counted at least 37 U.S. arrests tied to data-center protests this year, including a physics teacher from Kansas hauled out of a public meeting after clapping for an anti-data-center speaker. The crowd of those cuffed spans farmers, teachers, suburban parents, plus Zoomers and Boomers alike.

Cuffing season came early cause errybody in the club getting cuffed up!

Also, this is starting to cost real money: local opposition helped delay or block $130B worth of U.S. data-center projects in Q1 alone. Maybe it’s time we start referring to “the cloud” by its real name: the middle of the desert where nobody will complain about it. Or, y’know: actual space!

Here’s what happened in AI today:

• 😺 OpenClaw 2.0 rebuilt setup, memory, and shared agents.

• 📰 OpenAI bought tens of thousands of Macs for agents.

• 📰 G20 leaders were warned frontier AI could shake finance.

• 📰 DiDi began fully driverless R2 passenger trials.

• 🎓 Build a citation-grounded project brain with Gemini Notebook.

…and a whole lot more that you can read about here.

😺 OpenClaw 2.0 turns the DIY agent into a real platform

If OpenClaw, the personal AI agent tool always sounded like something that was too complicated or scary for you (because you needed a terminal, three API keys, and a Saturday afternoon to pull your hair out trying to learn SSH and Gateways and whatever else), there’s a new version out aimed at solving ALL THOSE problems.

The open-source personal agent just shipped v2026.8.1, aka OpenClaw 2.0, its largest update yet. The project says 933 contributors, including 569 first-timers, merged more than 16,000 pull requests into this release. May the vibes be with y’all.

Here’s what happened:

• Your first-run setup of OpenClaw 2.0 can actually reuse existing ChatGPT or Claude subscriptions, API keys, or local models.

• The browser app was rebuilt around ongoing conversations, dashboards, progress tracking, and interactive widgets.

• Shared cloud sessions can move work to paired devices or cloud workers, then hand the same session and context to another person.

• Memory adds conversation recall, background consolidation, and reusable-skill learning; Labs adds experimental Swarm and Fleet modes.

Okay lets talk Swarm vs. Fleet: Swarm mode is one OpenClaw task that spawns parallel subagents and collects their results. Fleet produces multiple isolated OpenClaw “cells,” each with its own Gateway, credentials, and state.

Think of the difference like hiring a temporary team vs. setting up separate offices for different departments. Remember offices with actual people in them??

How to try OpenClaw 2.0:

1. Open the official install guide. Mac, Linux, and WSL2 users can run the recommended installer; Windows users can use the signed Hub app or PowerShell installer.

2. Follow onboarding to choose your model and authentication, create a workspace, and install the Gateway as a background service.

3. Use the browser Control UI or connect a messaging channel like Telegram or Discord, then give your Claw one useful recurring workflow.

Already using OpenClaw 1.0? Back up ~/.openclaw, switch to stable with openclaw update --channel stable, then run openclaw doctor --fix. This release includes breaking migrations for OpenProse and older OpenAI routes.

And if all that’s confusing, just hand these instructions and links to your current agent to help walk you through it!

Why this matters: OpenClaw is trying to move personal agents from hobbyist toy with lots of technical plumbing to navigate towards out of the box software normal people can set up and run themselves.

The model under the shell of the claw can be OpenAI, Anthropic, Google, or your own local or cloud-based open model; OpenClaw 2.0 handles your memory, permissions, tools, channels, recurring work, and now, collaboration with others.

The test from here is stability. OpenClaw’s release notes change almost every layer, and its GitHub release explicitly tells existing users to back up before upgrading. Can OpenClaw’s early adopters survive the transition, and is 2.0 enough to get normal people into the Eclawsystem? I guess it depends on if YOU, dear reader, try this version out or not!

FROM OUR PARTNERS

Join us Sept 16: Where Private and Public Capital is Betting on AI   

Join Meera Oak, Senior Principal at Alumni Ventures, and Corey Noles, Editor of The Neuron, for a fireside conversation on the AI infrastructure landscape — where private and public markets are placing capital, which companies are building the picks-and-shovels of the AI era, and how accredited investors can position themselves for one of the most consequential venture opportunities of the decade.

Save my seat

🎓 AI Skill of the Day: Build a Project Brain That Cites Its Sources

Reader request: how do you turn your own docs into a useful project brain without training a model? Our answer: Use Gemini Notebook (formerly NotebookLM), which answers from the sources you add and gives inline citations.

1. Create one notebook per project and add the official project Docs, PDFs, sites, videos, or Sheets.

2. For each question, select only the sources that should count as truth.

3. Ask for the answer, citations, contradictions, and missing evidence before you act.

Answer only from the selected sources. Cite every factual claim. If the sources conflict or do not contain the answer, say so and list the missing evidence.

Have a specific skill you want to learn? Request it here.

FROM OUR PARTNERS

Upgrade your voice stack with Deepgram Flux TTS

Flux TTS processes entire conversations rather than isolated text strings, keeping agent speech natural across multi-turn calls. You get as low as ~80ms streaming latency and native interruption handling built directly into the engine.

Test it on your stack now.

📰 Around the Horn

• OpenAI said ChatGPT Ads hit a $1B annualized revenue run rate less than 200 days after launch, with tens of thousands of advertisers across 40+ countries.

• OpenAI reportedly bought tens of thousands of Macs for reinforcement learning and computer-use agent training, while Anthropic rents Mac capacity through AWS.

• SB Energy gave OpenAI warrants now worth about $5.5B to secure the company as an anchor tenant for its planned data centers.

• The Financial Stability Board warned G20 leaders that frontier AI cyberattacks could hit multiple financial firms at once and threaten global stability.

• Sony and Warner music publishers sued Anthropic, alleging it pirated tens of thousands of songs to train Claude.

• DiDi, the Chinese robotaxi company, began fully driverless R2 passenger trials in selected Beijing and Guangzhou zones, with rides bookable in the DiDi app.

• Cloudflare launched Adaptive Intelligence, a bot-defense engine that continuously creates short-lived rules from live traffic to make automated attacks harder to sustain.

🍪 Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 700K+ readers here!

1. *Outskill’s free 3-hour workshop shows you 15 AI tools and how to build an AI coworker that works 24/7 this Saturday. Save my free seat.

2. Fermion Research released Phonon-1, a 415 MB local English speech-recognition model that can transcribe an hour of audio in roughly 2.5 minutes on a laptop. Try it here.

3. fal released H3 Max, a post-trained MiniMax H3 variant that generates a five-second video in under three seconds in its tests. Try the playground.

4. Google made Gemini Omni 1.1 Flash generally available for conversational video generation and editing, including extensions, interpolation, and 4K output. Try it in AI Studio.

5. Operant Semantic Firewall reads intent across prompts, tool calls, code, and data movement, then allows, blocks, or redacts risky agent actions in real time.

😹 Monday Meme

New from The Neuron: AI Explained

This new episode gets wild: Noam Schwartz walks us through how hackers could “groom” an AI agent over time, then how a compromised agent could influence other agents almost like human radicalization.

We also get into why prompt injection may never be fully solved and why agent security gets way scarier once AI can actually take actions. Buckle up!!

Watch / listen: YouTube | Spotify | Apple Podcasts

This video was sponsored by Dell AI Factory with NVIDIA. Learn more.

P.S: We’re trying to hit 50K subscribers on YouTube this year. Click here to help!

A Cat’s Commentary

P.S: Love the newsletter, but only want to get it once per week? Don’t unsubscribe—update your preferences here.