Gemini 4, Meta's Muse Wonder, and Character.AI's Bankruptcy Sale

Claire: Meta's not far behind with its first real frontier model, and it remembers things about you across weeks.

Grant: And Character dot AI, the chatbot platform, just filed for bankruptcy, with Disney circling the wreckage.

Claire: Plus Shopify just let any AI agent shop its entire catalog. We've got five stories and a lot to unpack.

Grant: I'm Grant.

Claire: And I'm Claire. This is The Z Media Podcast, your daily AI news for marketers, Wednesday, October fourteenth, 2026.

Grant: Okay Claire, Google's flagship is finally out. Gemini 4 Argon is live today, worldwide, through the API and through paid Gemini plans.

Claire: So if you're on the free tier, or even Plus, you don't get the new model at all?

Grant: Right. The Decoder says free and Plus users only get Flash models under this new tier structure. Argon replaces the Pro class as Google's top series, but it's gated behind the pricier subscriptions.

Claire: And what about the API pricing, because that's what our listeners actually budget around.

Grant: This is the part worth circling. Intro pricing is two dollars per million input tokens and ten dollars per million output, with cached input ninety five percent cheaper. But Ars Technica is clear that rate doubles after November thirtieth, to four dollars and twenty dollars.

Claire: So this is basically a limited time offer to stress test the model cheap.

Grant: Exactly. Google says Argon ranks first on the Vals Index, which covers things like finance, coding, legal and tax work. But The Decoder notes it still trails Claude Opus 5.5 in independent overall rankings. So take the top ranking with a grain of salt, it's Google's benchmark.

Claire: What's the actual move for a content team this week?

Grant: Test it now on real workflows, document analysis, drafting, agent tasks, while the price is low. Measure your total token use before November thirtieth, because that's when the bill changes.

Claire: Alright, my turn. One day before Google's launch, Meta Superintelligence Labs released two models, Muse Wonder and Muse Wonder Full, on October twelfth. TechCrunch calls this Meta's first frontier class model.

Grant: Where can people actually use it?

Claire: Wonder is on the Meta Model API and in Muse Code, their coding tool. Wonder Full is locked to Muse Code subscribers only, it's not in the API at all, and there's no published price for it yet.

Grant: What's the pricing on regular Wonder, since that's the one people can actually build with?

Claire: The Decoder reports four dollars per million input tokens and twenty dollars per million output. Compare that to Muse Spark one point three, which was a dollar twenty five and four twenty five. So it's a real jump, almost three times the cost.

Grant: Is the performance jump worth that cost increase?

Claire: Artificial Analysis scores Wonder Full at fifty six and Wonder at fifty two on its Intelligence Index, up from forty eight for Spark. Biggest gains are in agentic coding. But Wired points out a lot of this comparison data comes from Meta's own testing, and benchmarks don't always match real world performance.

Grant: You mentioned agent memory earlier. What's actually new there?

Claire: Two features, Active Memory, which keeps key user info across days and weeks, and Active Focus, which keeps the model on task and lets it ask clarifying questions. Meta says Wonder will power its Muse personal agent too, starting with paying subscribers.

Grant: Any red flags for teams thinking about building agents on this?

Claire: Wired notes Meta hasn't said whether Wonder powered agents misbehaved in testing. And The Decoder found weaker factual knowledge scores than top rivals, though lower hallucination rates offset that somewhat. If accuracy matters for your content, run your own tests before switching.

Grant: Switching gears, here's a fun one. Google Play now lets you launch apps you built in Google AI Studio, the vibe coding tool, straight to your own Android phone. No public listing, no developer account.

Claire: So this is for personal use only, not something you publish for customers?

Grant: Right now, yes. The Verge says these private apps don't show up in search or listings, only the creator can use them. Ars Technica reports they land in a Personal Library on the device, and Google doesn't review them the way it reviews public apps.

Claire: Who gets access, and where?

Grant: It's starting in the U.S. for users eighteen and older, with Ars Technica noting a few other markets are included too. Sharing with a small group of friends and family is coming later, but Google hasn't set how big that group can be.

Claire: What's the actual use case for a marketing team here?

Grant: Think small internal tools. A campaign checklist, a content calendar viewer, an event check in app. You can build it, put it on your phone, no store listing needed. TechCrunch points out this used to push people toward web apps instead, since publishing or manual installs were such a hassle.

Claire: And if something takes off, can it graduate to a real public app?

Grant: Yes, Google's pitching it as a path. Apps that catch on can move to a public listing after the same trust and safety review other developers face, plus a developer profile. We'll put the links in the show notes if you want to try building one.

Claire: This next one is heavier. Character dot AI, the chatbot platform where people build and talk to custom characters, filed for Chapter eleven. TechCrunch reports assets and liabilities listed between fifty and a hundred million dollars.

Grant: Does the platform shut down immediately?

Claire: No, the company says it keeps running during a court supervised sale, and existing subscriptions continue. But CEO Colin Adams said if no workable bid shows up, the platform gets wound down entirely.

Grant: Is there a buyer lined up?

Claire: Disney has agreed to pay twenty million dollars for some assets, plus a non exclusive perpetual license to the technology. Ars Technica says the sale is expected to close late November or early December. The Decoder notes this is a steep fall, Character dot AI was once part of a two point seven billion dollar Google deal.

Grant: What about user data? That's what I'd worry about if I had a branded character on there.

Claire: Good instinct, because The Verge is asking exactly that question. Forty bipartisan state attorneys general filed a notice of interest, seeking age checks, limits on minors' use, crisis safeguards, and the ability for users to delete their data from any buyer. New York's Attorney General said any buyer needs to put children's safety first.

Grant: So what's the lesson for marketers who use consumer AI tools generally?

Claire: Platform risk is real, even for products with huge engagement. If you've built an audience, a partnership, or an experiment on a consumer AI platform, have an exit plan, and know how that vendor handles your data if ownership changes hands.

Grant: Last story, and it's a big one for e-commerce brands. Shopify announced Tuesday that shoppers can connect any AI agent to their Shop account and buy products, no partnership integration required, according to The Verge.

Claire: How big is the catalog we're talking about?

Grant: Over a billion listings, including items both on and off Shopify storefronts. Before this, the main route was a September partnership where Meta's Muse agent could browse and buy. Now there's no dedicated deal needed at all.

Claire: Does the agent just buy whatever it wants, or is there a human check?

Grant: Humans stay in the loop. Checkout runs through Shop Pay inside the conversation, and the buyer has to approve every purchase. Orders flow back into the Shop app. Shoppers can also set rules, a budget cap, or fast shipping only results.

Claire: So Shopify is calling this a personal shopper. What does it mean for how brands present products?

Grant: This is the real shift. If an agent is comparing products against a shopper's rules, then your product titles, descriptions, pricing clarity and shipping details are what the agent weighs, not a landing page a human reads.

Claire: Is Shopify alone in this, or is this part of a bigger fight?

Grant: The Verge notes Google and Amazon are investing in similar tools, and Amazon is already in an ongoing dispute with AI companies including Perplexity and now Meta's Muse over whether agents should be allowed to shop on a customer's behalf. Worth watching how merchants will even be able to measure agent driven orders.

Claire: Okay, pulling this all together. What's the one thing a marketing team should actually do this week?

Grant: I'd say run a real test on Gemini 4 Argon while the intro pricing lasts, two dollars and ten dollars per million tokens, before it doubles on November thirtieth. Put it on an actual task, not a toy example.

Claire: And pair that with a quick audit of your product listings for agent readiness, since Shopify just opened the door for agents to shop without a platform deal. Clear titles, clear pricing, clear shipping info, that's the groundwork now.

Grant: And if you've got anything built on Character dot AI, check your exit plan this week too, not after the sale closes.

Claire: That's a full plate for a Wednesday. Between model pricing shifting under your feet and agents starting to do the shopping, this is exactly the kind of thing that's hard to track manually every single day.

Grant: Which is the whole idea behind what we build at Z Media. We put together fully automated AI content and CRM systems that write, design, film, publish and even answer for your brand, so you're not chasing every pricing change or platform shift by hand.

Claire: If you want that kind of automation working for your own brand, talk with our team. The links are in the show notes, or call or WhatsApp us at four one nine, five zero zero, four eight eight two.

Grant: And if you just want the key points every day, subscribe to the AI news wire. That's it for today, we'll see you tomorrow.

Claire: Thanks for listening. This has been The Z Media Podcast.