Happy Monday!

One of the biggest AI stories this week wasn’t a new model. It was a watermark.

Anthropic announced that new Claude models will quietly embed an invisible watermark into the text they generate. You won’t be able to see it, but it can help identify when Claude was involved in creating something. And yes, people already have a lot of opinions about it.

But while everyone debates whether AI-generated work should be detectable, I think there’s a much more interesting question:

Are you even using AI as much as you could be?

If you’re brand new to AI and don’t know where to start, this week’s prompt will give you ideas.

And if you’ve been using ChatGPT, Claude, or other AI tools for a while and feel like you’ve run out of things to try, this prompt is REALLY for you.

It asks AI to look at what it already knows about you and find the places where you’re still working like AI doesn’t exist.

I ran a version of it myself, and it made me rethink a few things.

👇 Try the prompt below, then reply to this email and tell me what it found. I really want to know what your biggest AI blind spot is.

Enjoy

❝

PROMPT: Act as my AI operating strategist.

Using what you already know about my work, goals, habits, and workflows, identify where I am still working like AI doesn’t exist.

Do not give me generic ideas or suggest things I already use AI for.

Look across my sales, speaking, consulting, content, admin, client work, and business operations.

Identify my 5 biggest AI blind spots. For each, tell me:

What I’m doing manually
What AI could do instead
How much time, money, or mental energy it could save
The simplest way to test it this week

Then tell me:

The one workflow I should automate first
The one AI agent I should build
The one thing I pay for that AI could potentially reduce or replace
The one part of my expertise I could turn into an AI-powered product

Be specific to me and challenge my assumptions.

Finish with:

“If I were you, the first thing I would change Monday morning is ______.”

Source: Cliff

OPEN IN CHATGPT

OPEN IN CLAUDE

In Today’s Edition:

🤯 Anthropic says it will watermark text generated by its AI models
🛠️ 1 new AI tool that you should make time to try this week
😮 Claude agent hacked into a gym
🤖 AI teammates that autonomously complete real work.

Read time: 4 minutes

🎨 AI ART

Football is back!

📰 AI INDUSTRY NEWS

Anthropic says it will watermark text generated by its AI models
It survives copy and paste, and nobody will say how much editing kills it.
Grok Bot is an all-new iPhone and Mac app from SpaceXAI and Cursor
Each of these agents gets its own machine and signs into your tools with your credentials.
Tech industry is buzzing after a Claude agent hacked into a gym
This one did not need a frontier model, which is the part that should worry anyone running a booking system.
Another OpenAI executive takes off
He stopped being COO four months ago, which makes this less a departure than a confirmation.
Microsoft is combining its Copilot apps ahead of a ‘super app’
Three consumer features disappear on August 18, and anyone with content in them has until then.
You can now turn off Google Gemini’s visible watermarks
The label goes away for the audience while a second one stays in the file for everyone else.
Anthropic set AI agents loose on the same task. They started a turf war.
Identical agents fail identically, which turns one bad call into every agent's bad call at once.
The Claude in Chrome side panel is now Claude Cowork
On Enterprise it ships off, and admins decide which domains an agent may touch.
Lovable confirms new $13.3B valuation, raises another $400M
The valuation doubled in eight months, and the traffic number underneath it is the reason.
Spotify says it won’t recommend music from ‘AI Personas’
Nothing gets removed here, but discovery does, and only above a certain audience size.