Author: Evolving AI - Bewertung: 2x - Views:24
Meta just bought the fastest-growing AI agent company in history – and when you look at the numbers, the deal suddenly makes perfect sense. In this video, we break down why Meta spent around $1–2 billion on Manus, a China-founded, Singapore-based AI agent startup that hit $100M ARR in eight months, processed 147 trillion tokens, and spun up more than 80 million isolated Linux “virtual computers” in a single year. This isn’t another model story. Meta didn’t buy a brain – it bought a production-tested execution layer that can actually finish workflows: a planner that breaks big goals into steps, an executor that writes and runs Python inside sandboxes, and a verifier that checks if the task really got done. Manus doesn’t just predict the next word – it predicts the next action. We’ll walk through why Meta’s AI problem in 2025 wasn’t compute or models, but reliability: Llama 4 lagging GPT-5 and Claude on key benchmarks, the messy Meta AI rollout, and a stack that could chat but couldn’t consistently automate real business tasks. That’s why Meta first bought into Scale AI and hired Alexandr Wang as Chief AI Officer, then came back for Manus to plug the execution gap. You’ll see how Manus routes tasks between different models, caches successful scripts, and embodies “Situated Agency” – the idea that real intelligence comes from coupling models to tools, memory, sandboxes and verification loops. And we’ll look at what happens when this framework is wired directly into WhatsApp, Ray-Ban smart glasses, Quest, Instagram and Messenger, and eventually into Meta’s closed Avocado model. Foundation labs are the brains. Agent labs are the bodies. Meta just bought one of the best bodies on the market – and if you want the real story behind the world’s fastest-moving AI breakthroughs, make sure to like and subscribe to Evolving AI for daily coverage.
SOCIAL SHARE CARD GENERATOR