Good morning, AI enthusiasts. Zhipu AI just launched GLM-4.5, a unified model that combines reasoning, coding, and agent capabilities in a single 355-billion parameter system.
Could this push toward "one AI for everything" finally kill the era of specialized models?
In today's recap:
China's GLM-4.5 unifies reasoning, coding, and agents
Microsoft transforms Edge into AI web assistant
Anthropic caps Claude Code usage for power users
Meta hires OpenAI researcher as AI lab chief
5 new AI tools and more
ZAI
Zhipu AI drops GLM-4.5 with unified reasoning power
Source: Zhipu AI
Recaply: Chinese AI company Zhipu AI just launched GLM-4.5 and GLM-4.5-Air, new flagship models that combine reasoning, coding, and agent capabilities in a single system. Both models offer hybrid thinking modes for complex tasks.
Key details:
GLM-4.5 packs 355 billion total parameters with 32 billion active, while GLM-4.5-Air uses 106 billion total with 12 billion active parameters.
The models ranked 3rd and 6th overall across 12 benchmarks covering reasoning, coding, and agent tasks.
They feature both a thinking mode for complex problems and a non-thinking mode for instant responses.
Both models are available through the Z.ai platform with competitive API pricing starting at $0.2 per million input tokens.
Zhipu AI released open-source weights on HuggingFace and ModelScope for local deployment.
Why it matters: GLM-4.5 represents a big push toward one AI that's good at many things instead of separate AIs for different jobs. By combining reasoning, coding, and assistant features, Zhipu AI challenges the current approach of using different models for different tasks.
MICROSOFT
Edge browser becomes your AI web assistant with Copilot Mode
Source: Microsoft
Recaply: Microsoft just unveiled Copilot Mode for Edge browser, bringing AI features that predict what you need and handle web tasks automatically, changing how we browse the internet.
Key details:
Copilot Mode is a free experimental mode on Windows and Mac that you can turn on or off.
It reads across multiple browser tabs to understand your full research and give better suggestions.
Voice commands let you navigate and complete tasks without typing.
Smart help includes changing recipes, summarizing articles, and comparing prices.
Privacy settings let you control when AI can see your browsing.
Why it matters: Microsoft thinks AI assistants built into browsers will become normal for everyone. This challenges Google's search control by putting intelligence directly in the browser. Features like cross-tab research help and voice navigation could attract people who aren't tech experts and those who need accessibility support.
ANTHROPIC
Anthropic puts brakes on Claude Code with new weekly limits
Source: Anthropic
Recaply: Anthropic is rolling out new weekly rate limits for Claude Code starting August 28, targeting the small group of power users who run the AI coding tool continuously and those violating usage policies through account sharing.
Key details:
Less than 5% of users will hit the new weekly limits on Pro and Max plans.
Pro users get 40-80 hours of Sonnet 4 per week, and Max users get much more.
One user burned through tens of thousands of dollars in AI usage on a $200 monthly plan.
Claude Code has crashed seven times in the past month due to high demand.
Max users can buy extra time beyond limits at regular API prices.
Why it matters: This shows how hard it is for AI companies to balance popular features with the cost of running them. As coding tools become must-haves for developers, companies need to find ways to serve both light users and heavy users without crashing their systems.
META
Meta hires ex-OpenAI researcher as AI lab chief
Source: Alexandr Wang
Recaply: Meta CEO Mark Zuckerberg just announced that Shengjia Zhao, a former OpenAI researcher, will be Chief Scientist of Meta Superintelligence Labs, formalizing his leadership role in the company's new frontier AI research division.
Key details:
Zhao co-authored the original ChatGPT research paper and contributed to GPT-4, o1, and o3 models.
He co-founded Meta's superintelligence lab and has led scientific efforts since day one.
Zhao will work directly with Zuckerberg and chief AI officer Alexandr Wang to set research direction.
Yann LeCun remains Meta's chief AI scientist for FAIR, focusing on long-term research.
Zhao brings expertise in synthetic data generation and scaling paradigms from OpenAI.
Why it matters: This hire completes Meta's plan to recruit top AI talent from other companies. With expert researchers and billions spent on computers, Meta's AI team is now ready to compete directly with OpenAI and other leading AI companies.
NEWS
Beyond the big stories in AI today
Alibaba launched Wan2.2, the first open-source video generation model with cinematic control and mixture-of-experts architecture for complex motion.
Kling AI introduced Kling Lab, a collaborative workspace designed to streamline creative processes, currently in beta testing.
xAI is developing "Imagine," a feature for Grok that will create images and videos. It will be available soon in early access.
Google Chrome added AI-powered store summaries for US shoppers, generating reputation insights from review partners to improve purchase decisions.
Similarweb released Gen AI Intelligence, tracking AI brand visibility and competitor traffic to show how AI reshapes market discovery.
Google Research unveiled SensorLM, a foundation model trained on 60 million hours of wearable data for natural language sensor understanding.
TOOLS
Trending AI Tools
🎬 1of10 - Discover viral YouTube video ideas and thumbnails using outlier data
🚀 ManyChat - Add AI Steps to automate chat flows on Instagram, WhatsApp, TikTok
🎤 Letterly - Convert your voice into polished text for notes, emails, posts
🎥 Higgsfield - Generate personalized UGC-style videos and images with AI
🔊 Krisp - Remove background noise, transcribe meetings, even convert accents in real time
