🧠 Meta acquires Manus AI 🧠
Meta’s latest big-ticket move doubles down on agentic AI — and recurring revenue.

Meta* has agreed to acquire Manus AI, a Singapore-based startup that shot to fame this year with demos of an autonomous “general AI agent” handling real workflows (research, planning, analysis) with minimal prompting. While Meta hasn’t disclosed the price, multiple outlets report a multi‑billion‑dollar deal, and Meta says it will keep Manus operating independently while weaving its agents into Facebook, Instagram, WhatsApp, and Meta AI. Manus has also claimed millions of users and $100M+ in ARR, making it a rare “agent” product that appears to have reached meaningful paid adoption at speed. There’s a geopolitical wrinkle: Manus was originally founded by Chinese founders and later relocated to Singapore. Meta can now potentially place agents directly inside the world’s biggest consumer apps, redefining how AI assistants monetize — and how quickly they can scale.

📊 21% of YouTube's Feed Is Now 'AI Slop' 📊
New research exposes the rapid spread of low-quality AI content on YouTube.

Video editing company Kapwing has released a report revealing that over 21% of YouTube videos recommended to new users are low-effort AI-generated content designed to farm views. The study analyzed the first 500 recommendations shown to a fresh account, exposing the scale of automated content flooding the platform. The top-performing channel was India's Bandar Apna Dost, an animated monkey series with over 2 billion views and an estimated $4.25 million in yearly ad revenue. South Korea led in total slop viewership, followed by Pakistan and the United States.

🤖 Claude Code Now Writes All Its Own Code 🤖
Anthropic's agentic coding tool achieves full self-authorship milestone.

Anthropic's Claude Code creator Boris Cherny revealed that in the past month, 100% of contributions to the agentic coding tool were written by Claude Code itself. This milestone demonstrates advancing AI capability to autonomously develop and improve its own codebase without human intervention. The development raises important questions about AI self-improvement and the future role of human developers in maintaining AI tools. This matters for the competitive pressure in agentic coding, as Anthropic showcases progress in building reliable systems capable of complex software development tasks entirely on their own.

INTERESTING TO KNOW

💻 Z. ai’s GLM-4.7 Tops Coding Benchmarks 💻

Chinese AI startup Z. ai released GLM-4.7, a coding-focused model achieving 73.8% on SWE-bench—making it the first Chinese lab to break 70% on this real-world coding benchmark. The model seems to surpass open-source competitors including DeepSeek-V3.2, Kimi K2, and Claude Sonnet 4.5. The rollout is relevant for competitive pressure, as the Alibaba-backed startup expects to raise $300M next month following Hong Kong IPO hearings.

📜 China Regulates Human-Like AI 📜

China has issued draft regulations specifically targeting AI systems with human-like interaction capabilities. The new rules address concerns about AI addiction, emotional manipulation, and security risks posed by AI that mimics human behavior. This is unsual for China and could set a precedence on more regulations of chinese AI advancements, signaling a more ethical approach.