🦀 OpenClaw in the browser. Finally

Good Morning! Moonshot AI just pushed agentic workflows straight into the browser with Kimi Claw, turning kimi.com into a 24/7 AI agent workspace. Plus, you’ll learn how to access a massive open-source repository that shows how modern AI products are actually built.

Plus, in today’s AI newsletter:

• How to Access Real Production-Grade AI Architectures

• Moonshot AI Brings Kimi Claw to the Browser

• OpenClaw Creator Joins OpenAI

• OpenAI model Tackles Frontier Math Problems

• 4 new AI tools worth trying

AI MODELS

⚙️ Moonshot AI Brings Kimi Claw to the Browser

Moonshot AI has launched Kimi Claw, a native, cloud-based OpenClaw environment on kimi.com built for developers and data-heavy agent workflows.

• ClawHub gives access to 5,000+ community-built skills that can be discovered, chained, and orchestrated instantly

• 40GB persistent cloud storage enables large datasets, RAG workflows, and long-term context across sessions

• Pro-grade search pulls live data (e.g., finance, technical sources) to reduce hallucinations and stay up to date

• BYOC + app bridging lets developers connect custom OpenClaw setups and deploy agents to platforms like Telegram

Kimi Claw shifts agentic AI from local setups to a managed, always-on browser environment. By combining massive skills, real-time data, and serious storage, Moonshot AI is positioning the browser tab as a full-fledged AI development and automation platform.

AI NEWS

🦾 OpenClaw Creator Joins OpenAI

Peter Steinberger, the developer behind OpenClaw, an AI assistant built to manage real-world tasks, has officially joined OpenAI. The move comes as OpenAI doubles down on next-gen personal agents.

• OpenClaw went viral for handling calendars, bookings, and autonomous actions across platforms

• Steinberger says joining OpenAI is the fastest way to bring this tech to everyone

• Sam Altman says Steinberger will help drive the next generation of personal AI agents

• OpenClaw will remain open source, supported by OpenAI via a foundation

This shows OpenAI is serious about agents that do, not just chat. Bringing in builders like Steinberger hints that hands-on, autonomous AI assistants are about to move from demos to daily life.

AI MODELS

🧠 OpenAI model Tackles Frontier Math Problems

An internal OpenAI model, running with minimal human supervision, attempted the “First Proof” challenge, ten open, frontier-level math problems, and delivered promising solutions in just one week.

• The model produced solutions to most problems, with experts believing at least 6 of 10 are likely correct

• No proof hints were given; the model worked largely autonomously, with limited expansion requests

• The sprint was run in one week using a single in-training model, not a polished evaluation setup

• Full solution attempts will be published later, following author guidelines

AI models are starting to engage directly with unsolved research-grade mathematics, which may become one of the most meaningful ways to evaluate true next-generation AI capabilities.

HOW TO AI

🗂️ How to Access Real Production-Grade AI Architectures

In this tutorial, you’ll learn how to access a massive open-source repository that shows how modern AI products are actually built.

🧰 Who is This For

• Creators who want luxury, cinematic visuals without pro equipment

• Photographers looking to enhance images instantly

• Designers and marketers who need premium-looking visuals fast

• Anyone who wants photorealistic results with minimal effort

STEP 1: Access the Repository

First, you need to reach the source of everything. Open your browser and go here.

When the page loads, you’ll see a long list of folders and files. Think of this page as a table of contents for modern AI engineering. Each folder represents a complete, standalone project. For example, folders like agent-with-mcp-memory, DeepSeek-finetuning, or RAG-related directories are full implementations, not snippets.

You’re now looking at how real AI products are structured behind the scenes.

STEP 2: Find the “Green Key” (The Code Button)

Next, you need a way to take this code off GitHub and onto your machine.

Look near the top-right area of the file list. You’ll see a bright green button labeled “<> Code”. Click on it.

This button is important because it unlocks all download options, whether you want the code synced for development or just want to explore it locally.

STEP 3: Download the Code (Two Ways)

At this point, you have two ways to proceed. Choose based on how you plan to use the repository.

If you’re comfortable using the terminal and want to work with this long-term, copy the HTTPS link shown in the menu. Then open your terminal and run the git clone command using that link. This creates a local folder that you can update later as the repository evolves.

If you just want to explore quickly, click “Download ZIP” instead. This gives you a compressed file with every project included. No setup, no commands.

STEP 4: Verify You Have the “Blueprints”

If you cloned the repository, open your terminal, move into the ai-engineering-hub folder, and list the contents. You should see multiple project directories.

If you downloaded the ZIP, go to your Downloads folder, unzip the file, and open the extracted directory.

The key check is simple. Look for a folder named agent-with-mcp-memory. If you see it, you now have access to real templates for building AI agents with memory, tool usage, and structured reasoning.

Apple tweaked its Maps app for Tesla before the EV maker's planned addition of CarPlay; low-cost MacBook will get an aluminum chassis in playful colors.

As AI and agents are adopted to accelerate development, cognitive load and cognitive debt are likely to become bigger threats to developers than technical debt.

Following Disney, Paramount sent a cease-and-desist letter to ByteDance, alleging its AI-generated Seedance videos and Seedream images infringed Paramount's IP.

Global installed base of active smartphones grew 2% in 2025; eight OEMs had 200M+ active devices each, and nearly one in four active smartphones is an iPhone.

🦀 KimiClaw: Use open OpenClaw in the browser

🧑‍💻 Claude Opus 4.6: Built for long, serious work in coding and research

⚙️ GPT-5.3 Codex: Faster coding model with strong reasoning

🎨 Qwen-Image-2.0: Alibaba’s all-in-one image generation and editing model

THAT’S IT FOR TODAY

Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!

See you tomorrow :)

- Dr. Alvaro Cintas