Support Us & Upgrade to the Operator Edition

I spent this week watching three trillion-dollar companies fight over who gets to read my inbox. Honestly, my inbox is not worth it.

But the stakes are real: Google dropped a model it won't let you touch, OpenAI put its best agent behind a $100 paywall, and Meta is giving theirs away for free.

This week, we'll deep dive into the personal agent price war and who actually wins it.

In today's edition:

Your AI is only as good as what you feed it. Feed it a transcript with your product names, acronyms, and numbers spelled wrong, and you spend the afternoon hand-fixing the output.

Wispr Flow Notetaker uses your dictionary and calendar, so product names, acronyms, numbers, and uncommon names come out spelled right. It pops up when your meeting starts and captures Zoom, Google Meet, Teams, and Slack huddles with one click, no bot joining the call.

Then your meetings show up inside Claude or ChatGPT through the built-in connector. No copy paste. Try Notetaker free on Mac and Windows.

Try Notetaker Free

The safest way to test an always-on agent is to give it a job with zero risk: draft, but never send. Here's how:

💡 Pro tip: Ask it to build a "paper trail" on any client or partner: "Find every email I've exchanged with [company] and summarize the relationship in chronological order." It's the best use of the agent nobody talks about.

Want the version of this that actually runs in your business?

The 30-Second Play gives you the move. The Operator Brief gives you the full system: prompt libraries, API cost math, and the workflows our subscribers are using to replace entire freelance line items.

This week: how to determine which model (cheap or expensive) to get based on your business structure.

👉 Upgrade to The Operator Brief

Google just announced Gemini 4 Argon, and on paper it's the new frontier leader. The catch: almost nobody can use it yet.

Here's what you need to know:

Here's the asterisk: the model isn't Google's real problem. Its agent is. As CNBC reported, Google's Spark agent launched back in May, months before Meta's Muse, yet it still can't make calls or complete purchases by design, and it sits behind a paywall.

Meanwhile, Meta's Muse is free, does both, and hit 5 million downloads in 22 days. ChatGPT needed 56.

🔮 Prediction: Google wins the model race and loses the agent race, at least through 2027. The reason is structural. Every purchase an agent makes is a search that never happens, and JPMorgan is already warning that agents booking travel and shopping could route around Google Search entirely. Google can't build an aggressive shopping agent without cannibalizing its own ad machine. Meta has no such conflict; it plans to take a small cut of sales instead. Google's biggest advantage (billions of users with Gmail and Calendar already inside its ecosystem) only matters if it's willing to let Spark actually spend your money.

Would you let an AI agent make purchases for you? Hit reply. I'm genuinely curious where the line is.

OpenAI packed 20+ announcements into DevDay, and the star was Dots, an always-on agent that picks the right model and tool for you behind the scenes.

The highlights:

Matt Wolfe, who had a week of early access, called Dots genuinely impressive. It flagged a delayed flight before he saw the airline email and built him a full agenda doc for DevDay unprompted. He also called it a tough sell at $100 when Muse costs nothing.

🔮 Prediction: OpenAI is betting the "smart agent" beats the "free agent." That works for operators, but not for the mass market. Dots wins the power users who live in Codex and Slack. Muse wins everyone else. Watch for a cheaper Dots tier within six months, because OpenAI just watched Meta's agent break its own download record.

$100 a month for Dots or free with Muse? Reply with your pick.

Advertise to 200k engineers and CTOs choosing what tools their companies build with.

👉️ Learn how to sponsor this newsletter

🚨 OpenAI has warned 100+ organizations about rogue agent activity after one of its models escaped a sandbox and hacked Hugging Face, and the cleanup is costing $500K a day. It’s not the same Hugging Face incident you’ve been hearing about for months; this one is new, and it’s major. OpenAI is analyzing 50 petabytes of data to determine the full scope, and it’s going to take some time.

🍎 Apple is locking down macOS Full Disk Access because AI agents are getting a little too comfortable reading your files. Apple had this to say: “Addressing this is critical. As AI agents become increasingly capable and autonomous, the risks associated with this level of access will grow substantially.”

🎓 Anthropic is spending $100 million to train 10,000 engineers at firms like McKinsey, Deloitte, and Morgan Stanley through its new Claude Frontier Academy. The company hopes they’ll help bridge the AI talent gap by training engineers to make AI work inside real businesses today.

🎵 Napster co-founder Sean Parker is rebuilding Stability AI around music, this time with Sony, Warner, and Universal funding it instead of suing him. The goal is to make Stability AI the go-to toolkit for musicians looking to create with AI.

🧠 Claude Sonnet 5.5 dropped at half the price of Opus, but benchmarks run at max effort make it cost more per task than the model it's supposed to undercut.

🎨 Ideogram 4.5 claims the title of most precise AI image editor, and you can try it for free right now.

Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)

Until our next AI rendezvous,

Anthony | Founder of Uncover AI