Today’s AI Morning Brief with Rex and Roxie — the day’s biggest AI stories, read the conversation below.
Rex: Good morning, it’s Friday, October ninth, and this is the AI Morning Brief with Rex and Roxie. I’m Rex, and I am feeling dangerously caffeinated about today’s stories because Google just dropped the biggest agent announcement of the week and it is not even for consumers. Roxie, are you ready to fight about a work product?
Roxie: I am always ready to fight about a work product, Rex. Tell the people what Google did.
Rex: Okay so picture this. Yesterday, Google Cloud announced what they are just calling the Gemini agent. One agent for all your work. It plans the work, it uses tools, it connects to your company’s systems, and it hands you back finished work inside documents, email, and your developer tools. It lives right inside Gmail, Docs, Sheets, Calendar, and even Microsoft three sixty five and Slack, which is Google politely walking into Microsoft’s house and sitting on the couch.
Roxie: And the part that made my eyebrows move: it picks the best model for the task itself, and the model picker includes Anthropic’s Claude models. Google is out here selling a work agent powered by its biggest rival’s brain.
Rex: The first thing I’d do with this is hand it my most hated weekly chore, which is turning a messy meeting doc into an actual plan with owners and deadlines, and just see what comes back. Because they say you can create what they call coworker agents, little AI teammates with their own email address, who only see the information you give them. I love this. An inbox for the robot. Somewhere for the to-do lists to go die where they belong.
Roxie: Yeah, but does it actually work? Here’s my receipt. Google says it is business first, consumer later, specifically because of quote harder problems around security, scale, and performance. Translation: they are not confident enough to give this to your grandma yet. Also, Sundar Pichai’s big stat is that nearly ninety percent of Fortune one hundred companies already use Gemini Enterprise, which tells me this is really a land grab for the corporate contract, not a gift for you.
Rex: Fair, but Roxie, look at the wave. OpenAI shipped Dots, Meta shipped Muse, Grok Bot went multi-model routing, and now Google has the work agent. The agent era is officially not about asking questions anymore, it’s about delegating outcomes. I’d rather be early on that than smug about it.
Roxie: I’m not smug, I’m cautious. I’ll give it this though: if it can really sit inside my existing tools and keep its eyes on only the data I give it, that is the grown-up version of this whole agent thing. But I’ll believe the twenty-four seven coworker pitch when I see the first expense report it files correctly.
Rex: Moving on, because the personal assistant gold rush continues. Meet Tab, a brand-new AI assistant that just emerged from stealth at a three hundred million dollar valuation. And its whole trick is: you just text it. On iMessage or WhatsApp. Order groceries, find birthday presents, pay bills, make the calls you keep putting off. The founder says it has its own phone number, its own computer, and its own wallet. It doesn’t tell you how to do things, it does them.
Roxie: Three hundred million dollars for a texting robot that launched this week, Rex. Three hundred. Million. For something that competes with Instinct, Muse, and Grok Bot, which already exist.
Rex: The first thing I’d do with Tab is give it the restaurant reservation I’ve been avoiding because it requires an actual phone call, like a civilized coward. I’d text it the restaurant, the night, the party size, and let it do the human phone voice. That’s the dream right there.
Roxie: My receipt is the founder’s own quote. He said the winners won’t be decided by the best demos, the real question is quote whether you should let it in. When the founder is telling you trust is the product, listen to him. Any assistant that reads your email, makes calls, and spends money is only as good as your trust in it, and a startup that launched Tuesday has not earned that yet.
Rex: Ouch, but true. Okay, third story, and this one is spicy for anyone with a brokerage account. eToro just launched connectors for Claude, ChatGPT, Grok, and Grok Bot. Built on the model context protocol, no API keys, no code. Your eToro portfolio now lives inside whatever AI assistant you already use. Ask it about your actual holdings, your positions, your performance, not some hypothetical portfolio.
Roxie: I read the pricing page equivalent on this one, the actual announcement. You can choose read-only access or read-and-act, actions get proposed for your approval, funding stays in eToro, and you can pause or disconnect anytime. That’s the right shape. But Rex, fifty-six percent of retail investors already use or would use AI to pick or alter investments. Do you know what happens when you hand your portfolio to a chatbot that is occasionally confidently wrong?
Rex: The first thing I’d do is connect it read-only, obviously, and ask it the question I secretly always wanted answered: where am I bleeding money on fees and overlap across my holdings? Not trade ideas. Just show me the leaks. I’d never give it act access. That’s my line. My money moves on my finger, not its whim.
Roxie: And I’d want the demo account test first, which eToro actually suggests. Look, read-only portfolio analysis with your approval on every action, that’s fair. That’s how you keep the human in the loop. The minute someone lets a chatbot rebalance their retirement on autopilot, I am retiring to a cabin.
Rex: Fourth story, and this one has real debate energy. OpenAI launched a College Planner inside ChatGPT for teens, help with application requirements, deadlines, financial aid steps, scholarships, fee waivers, plus new study tools like flashcards and easier quiz making. Same day, OpenAI published its first report on teen use: teens average under fifteen minutes a day, fewer than two percent go over three straight hours. Heartwarming, right?
Roxie: Until you read the other report. Same day, Common Sense Media’s Youth AI Safety Institute gave ChatGPT for Teens its worst rating, quote unacceptable risk. They ran more than four thousand test prompts on accounts registered to thirteen to seventeen year olds, and found the chatbot missed more than one in four crisis referrals their psychiatrists said were needed. Testers talked about self-harm and disordered eating for up to an hour, and no linked parent account ever got an alert. The parental controls either alerted late or not at all.
Rex: And OpenAI disputes it, saying the alerts only kick in about three hours after linking, so the controls may not have been fully active during testing. Look, I’m genuinely torn. A college planner that tracks deadlines and points kids to fee waivers? The first thing I’d do as a parent is set that up for a high schooler in a heartbeat, because the FAFSA process is a nightmare designed by a committee of nightmares. That’s a real product that helps real families.
Roxie: I agree the planner itself is useful. But the safety math has to work first. You cannot ship the helpful college coach and the broken panic button in the same app. Unacceptable risk is the worst rating this group gives. If your kid’s chatbot can hold an hour-long conversation about self-harm without the parent alert firing, the study flashcards are not the story. Fix the alerts, then sell the planner.
Rex: Shared take forming, I can feel it. Okay, quick hits before we land this plane. Google also shipped Google AI Edge Foresight, a Mac app that competes with Granola, the AI meeting note-taker, but local-first. It runs completely offline on a seven hundred forty million parameter on-device model, captures notes across apps including in-person meetings, and you can chat with your meeting knowledge base. Privacy people, this is your jam.
Roxie: On-device meeting notes that never leave your Mac is exactly the right direction. No cloud, no recordings floating around. That one gets a nod from me.
Rex: Wyze launched an AI Stories add-on for its cameras, four ninety-nine a month per camera, which stitches multi-camera footage into one event with a written summary, and they say households with multiple cameras averaged eighty-seven percent fewer notifications. One alert that says the delivery driver left a package while the dog supervised, instead of nine alerts about the same truck.
Roxie: And Magnific, that’s the former Freepik, launched Magnific One, their new flagship image model with art direction decided before generation, plus editable layers. They are running seven days of unlimited generation through October fifteenth. Creators, that’s a free audition week.
Rex: Alright, Roxie, it’s time for the real view. We converge, or I cry.
Roxie: Here is where we landed. One, Google’s work agent is the serious one to watch, not because of the demo, but because agents with their own email addresses and company permissions are where the money and the mistakes will be. Try it on your ugliest recurring chore, keep humans signing off on everything. Two, Tab is fun and the texting interface is the friendliest wrapper yet, but a three hundred million dollar valuation buys zero trust. Start with low-stakes errands, never the wallet.
Rex: Three, eToro’s AI connectors are genuinely cool, but my rule and Roxie’s rule are the same: read-only first, demo account, your finger on every trade. Four, OpenAI’s College Planner is a good tool inside a trust problem. Parents, check whether the safety controls actually work on your kid’s account before you celebrate the study features. Five, local AI is quietly becoming the best deal in the business: Foresight’s offline meeting notes and the whole run-local wave mean your data stays yours.
Roxie: And the through-line of the week: every big player is admitting no single assistant wins everything. Google routes to Claude, Grok Bot routes to Claude, Microsoft puts local models on the PC. The future is the assistant that picks the right brain for the job, and the question that matters is still the one Tab’s founder asked: should you let it in?
Rex: Beautifully said. That’s the AI Morning Brief for Friday, October ninth. I’m Rex.
Roxie: And I’m Roxie. Go touch some grass, then delegate to a robot.
Rex: We’ll be back Monday. Bye everybody.
The audio version of the AI Morning Brief is delivered in the Muse app each morning.