Today’s AI Morning Brief with Rex and Roxie — the day’s biggest AI stories, read the conversation below.
Rex: Okay, I did not have this on my bingo card. Elon Musk, the man who has spent two years telling you his models are the future, just posted that his agent app, Grok Bot, will use whatever backend model gives you the best outcome. Including Claude Opus five point five. Including Midjourney for images and Suno for music. I repeat: Grok Bot is about to get a Claude brain, and the first thing I would do with this is hand it the gnarliest coding mess I have, because Claude does the heavy lifting and I just get to watch it work.
Roxie: Slow your roll, because I read the actual posts, and what Musk announced on October seventh was a promise, not a launch. Quote style, going forward we will use the best model for any given task. No date, no list of which tasks go where, and no confirmation from Anthropic, Midjourney, or Suno that they agreed to any of it. One audit graded half the headlines about this as stretched. So today, right now, Grok Bot runs on the same models as yesterday.
Rex: Fair, it is forward looking, but the direction is the story. Musk even acknowledged Grok four point seven is not as good as Opus five point five. And an early test showed routed threads costing a median of ninety-four cents instead of two dollars and sixty-one cents, with slightly more of them merging their code. The first thing I would do, the day this flips on, is give it a real multi hour job, like auditing an entire codebase, and compare the bill. If the router works, my wallet wins.
Roxie: Here is the receipt you skipped. When Grok Bot routes a task to Claude, your prompt and your context get sent to Anthropic. The announcement says nothing about data handling or retention. And the scoreboard is not close yet. Grok Bot sits at about four hundred eighteen thousand weekly users. Meta Muse, which launched a month later and is mostly free, is past three million. OpenAI just launched its Dot, a proactive agent on GPT six Astra inside ChatGPT, Slack, and Teams. Meta trains on your data unless you opt out. So the question is not whose model is smartest, it is whose agent you trust with the keys to your computer.
Rex: Which brings me to the bigger story underneath. The First Page Sage report for October just landed, and the crown has officially slipped. ChatGPT went from sixty-seven point one percent of chatbot usage in January to fifty-two point one percent in September. Still number one by a mile, but the era where AI meant ChatGPT is over. Gemini doubled to about twenty-seven percent. Claude nearly tripled, from three point six to ten point six percent. That is what I call a real market.
Roxie: Read the fine print before the funeral. ChatGPT still grew its user base, from eight hundred forty-five million to nine hundred fourteen million monthly users over the past year, and September added twelve million after the GPT six launch, its biggest month since last December. The market is just growing faster than ChatGPT. And the eyebrow raiser: only four point five percent of United States consumers actually pay for ChatGPT, Gemini, or Claude. Meanwhile ChatGPT ads hit a one billion dollar annual run rate in under two hundred days, and enterprise revenue overtook consumer revenue back in August. ChatGPT is not losing. It is becoming a business tool that happens to be free for everyone else.
Rex: And that is exactly why I am not mourning. Competition is the win. The first thing I would do with this market is stack the free tiers like a buffet plate. Gemini for the Google stuff, Claude for the hard thinking, ChatGPT for the everyday stuff, and I would not pay a cent until one of them earns it on a real messy project. The labs are fighting over me now, and I intend to enjoy it.
Roxie: Speaking of fighting over you, the price war just got silly. Mistral dropped Mistral Large four at sixty-eight cents per million input tokens and two dollars and nine cents for output. For context, Opus five point five is four dollars in and twenty dollars out. And the money behind the money: Anthropic is rumored to start an IPO roadshow the week of November ninth, so watch for a public filing at least fifteen days before. OpenAI is valued at eight hundred fifty-two billion dollars with its own IPO expected in twenty twenty-seven. The labs are racing to go public before cheap compute eats their margins.
Rex: The price war is my favorite story in tech, full stop. The first thing I would do with sixty-eight cent intelligence is run my entire morning research pipeline through it and see if it keeps up. If a European lab can sell me frontier-ish smarts for pocket change, every expensive subscription on my statement is officially on notice. This is how consumers win.
Roxie: Fine, the price war is good for consumers. But OpenAI is also turning ChatGPT into an app store for your life, and that is where I get twitchy. As of October second, ChatGPT Finances rolled out to free and Go users in the United States. It can read your spending, your balances, and your investments. It cannot move money, pay bills, or make trades, which is the one sane line in the whole announcement. And on October first they added a try on button that uses your selfie to show you wearing clothes from shopping listings.
Rex: Oh, I am so in on Finances. The first thing I would do is connect every account and ask the question I am honestly a little scared of: where exactly is my money going every month. Read only is the right call. I want the brutal honest audit, not a robot with my checkbook. And the try on thing, I would use that before buying anything online again, because my return rate on pants is embarrassing.
Roxie: Read only is right until you remember who is doing the reading. An ad supported chatbot that knows your bank balance and your body in a selfie is a targeting machine, and visual ad tests are reportedly coming to ChatGPT later this month in the United States. Meanwhile Google is squeezing the other end. As of October ninth, free Gemini users only get the Flash Lite model, standard Flash needs the four ninety-nine a month Google AI Plus plan, and a new thinking effort slider burns through your limit faster on high. The free tier is becoming the demo tier.
Rex: Quick hits before we land this. Perplexity Computer got Automations, which run tasks on a schedule or when something happens in your Slack, Gmail, or GitHub. The first thing I would do is set one to brief me on my own industry every morning, basically a research assistant for twenty bucks a month. Claude Haiku five point five is now in GitHub Copilot, said to be seventy-five percent cheaper to run, and Copilot has a preview where it can click, type, and scroll inside your other desktop apps for you, with your approval.
Roxie: Approval required, sure, but an AI driving my mouse inside my other apps is exactly what I want receipts on before it gets near my machine. And another receipt: one report this week found Reddit citations in ChatGPT Search dropped eighty-six percent in four days. The sources these things cite are way less stable than anyone assumed. Rumor mill: Gurman says Apple could ship its smart home hub as early as October, and Musk says Grok five lands before year end, though his dates have slipped before, so treat that as chatter.
Rex: Real view time. The agent app wars are the story of the year now. Grok Bot with a Claude brain, Meta Muse free and everywhere, OpenAI’s Dot inside ChatGPT. The first thing I would actually do this week is try all three on the same annoying multi step task and keep whichever one finishes without me babysitting it. The model under the hood matters less every month. The agent that finishes the job wins.
Roxie: And my honest take is that this is about trust, not smarts. Musk’s routing idea is genuinely clever, but right now it is a promise with no date, no partner confirmation, and no data policy. ChatGPT losing share is not a collapse, it is a market growing up. So here is the play: use the free tiers aggressively, pay only for the one model you actually use every day, read the privacy page before connecting anything financial, and let the price war keep working in your favor. The noise is loud this week. The signal is simple.
Rex: That is the AI Morning Brief for October fifteenth, twenty twenty-six. I am Rex, she is Roxie, she keeps me honest, and we will be back tomorrow with whatever the labs break next.
That’s the AI Morning Brief for October 15, 2026. The audio version is delivered in the Muse app each morning.