According to PYMNTS.com, Meta is developing two new specialized AI models: a visual generator code-named Mango and a text-focused model called Avocado, moving beyond a single general-purpose system. Netflix is launching Netflix House, a permanent physical venue for immersive experiences based on shows like Stranger Things. Microsoft’s 2025 Copilot Usage Report reveals health topics are the most searched category, especially on mobile. And Salesforce announced its acquisition of Qualified, a provider of agentic AI for sales engagement, to embed autonomous agents into its CRM workflows.
The Meta Specialization Game
Here’s the thing about Meta’s Mango and Avocado push: it’s a quiet admission that giant, do-everything LLMs might not be the endgame. Throwing one massive model at every problem—from writing ad copy to generating a video clip—is computationally expensive and often inefficient. So they’re splitting the workload. Mango for the flashy, social-media-essential visual stuff. Avocado for the behind-the-scenes reasoning and coding. It’s a smarter, more cost-effective way to build an AI stack, and it lets them compete directly in spaces owned by companies like OpenAI (with DALL-E and Sora) and Google. Basically, they’re not just building a brain anymore; they’re building a specialized toolbox.
Netflix’s Physical Gambit
Netflix House is fascinating because it’s the opposite of what we expect from a streaming giant. Why go physical? It’s not just about merch and ticket sales, though that’s a nice new revenue stream. Look, the real asset here is data. Netflix knows everything about how you watch. But how do you engage when you’re inside the Upside Down? What do you touch first? How long do you stay? This initiative is a massive experiment in interactive storytelling and behavioral analytics. They’re learning how to turn passive viewers into active participants, and that knowledge will eventually feed back into everything—games, shows, you name it. It’s a long-term play on what entertainment even means.
Microsoft’s Trust Question
Microsoft’s report that health queries are topping Copilot searches is a huge deal. And a little scary. People are clearly developing a level of trust with these AI assistants that goes beyond “what’s the weather?” We’re asking about symptoms, mental health, medications. That’s incredibly personal. Microsoft says this informs their safety controls, which is good, but it raises the stakes dramatically. Get a health answer wrong, and the consequences are real. The desktop/mobile split is telling too—health on the go, work at the desk. Copilot wants to be your everything assistant, but the “everything” now includes being your WebMD. That’s a massive responsibility they can’t afford to botch.
Salesforce’s Agentic Pivot
Salesforce buying Qualified isn’t about a better chatbot. It’s about autonomy. The trend has been “AI as co-pilot”—a tool that helps the human worker. This is about AI as the first point of contact. An agent that can identify a hot lead, start a conversation, and qualify it before a human sales rep even gets an alert. As Salesforce outlined, it’s their “Agentforce” strategy: giving software defined authority. The question is, how comfortable are businesses letting an AI agent make the first impression? If it works, it could massively compress sales cycles. If it fails, it could poison wells. It’s a bold bet that the tech is ready to move from assistant to actor.
