PLUS: Google Search gets a Gemini brain, NVIDIA’s new weather AI, and the AI chip gold rush

Good morning, AI enthusiast.

A viral new AI assistant is capturing the attention of the developer community by building its own tools on the fly. Clawdbot is a self-hosted agent that runs on your personal machine, offering a glimpse into a future of highly personalized AI.

While the agent offers powerful control over personal data and workflows, its rapid adoption has also exposed significant security vulnerabilities. The trend raises a key question: how do we balance the immense power of local AI with the critical need for security?

In today’s AI recap:

• The viral AI assistant that builds its own tools
• Google Search gets a Gemini brain
• NVIDIA’s new open-source weather AI
• The surging demand for AI memory chips

The Rise of Clawdbot

The Recap: A viral, open-source AI assistant named Clawdbot is capturing the attention of developers. It’s a self-hosted agent that runs on your own machine, giving you a personal AI that can manage tasks, access your local files, and even build its own new skills.

Unpacked:

• Clawdbot acts less like a simple tool and more like an AI teammate; you can ask it to create new capabilities on the fly, from setting up automations that replace Zapier to adding support for new AI models and services, showing what the future of a personal AI assistant could look like.

• Because it runs on your computer (Mac, Windows, or Linux), you have full control over your data and context, unlike with closed-off, cloud-based assistants. Its memory and skills are stored as simple text files on your machine.

• This power comes with a major caveat, as its rapid adoption has exposed significant security risks for users who run it on unprotected servers. The underlying protocol it uses was shipped without mandatory authentication, leaving many instances vulnerable.

Bottom line: Clawdbot offers a compelling glimpse into a future of highly personalized and capable AI agents that are deeply integrated into our workflows. Its popularity also serves as a crucial reminder that security must be built into these powerful systems from day one.

Google Search Gets a Gemini Brain

The Recap: Google is upgrading Search globally by making its powerful Gemini 3 model the new default for AI Overviews. This change allows users to ask more complex questions and get higher-quality AI responses directly on the results page.

Unpacked:

• The upgrade makes Gemini 3 the new engine for AI Overviews, providing users with best-in-class AI-generated summaries for their queries.
• Users can now ask follow-up questions directly from an AI Overview, seamlessly transitioning into a conversational “AI Mode” that maintains context.
• This creates a more fluid experience, offering a quick snapshot for simple searches and a deeper conversational dive for more complex topics.

Bottom line: This move signals a significant shift for the world’s most popular search engine, integrating generative AI more deeply into the core user experience. It aims to transform how we find information, moving from simple keyword results to dynamic, contextual conversations.

NVIDIA’s AI Weather Forecaster

The Recap: NVIDIA has launched Earth-2, a family of open and accelerated AI models designed to make high-accuracy weather and climate prediction faster and more accessible worldwide.

Unpacked:

• Earth-2 moves away from traditional physics-based models that require supercomputers, significantly cutting computational time and costs for organizations.
• The new models include Earth-2 Medium Range for 15-day global predictions and Earth-2 Nowcasting, a generative AI for high-resolution local storm forecasts just minutes ahead.
• Early adopters are already seeing major benefits, with the Israel Meteorological Service reporting a 90% reduction in compute time compared to its classic models.

Bottom line: Making these powerful weather AI tools open enables more industries to develop customized forecasting for their specific needs, from energy grid management to financial risk assessment. This open approach accelerates innovation and collaboration in understanding and predicting complex climate patterns.

The AI Gold Rush for Chips

The Recap: Chipmaker SK Hynix reported record-breaking profits, more than doubling its operating earnings. The boom is fueled by intense, surging demand for the high-bandwidth memory (HBM) chips that power AI data centers.

Unpacked:

• The numbers are staggering, with the company’s operating profit surging 137% in the December quarter compared to the previous year as HBM revenue more than doubled in 2025.
• A key factor in this success is a major partnership with Microsoft, where SK Hynix is the reported exclusive supplier of HBM for the tech giant’s new Maia 200 AI processor.
• This reflects a massive industry trend where demand for specialized AI memory chips far outpaces current supply, leading to shortages and surging prices that are expected to last into next year.

Bottom line: The incredible demand for specialized AI hardware creates a massive opportunity for the chipmakers who can deliver. Expect this high-stakes competition to accelerate innovation and drive the next wave of AI capabilities.

The Shortlist

Anthropic published Claude’s Constitution, a foundational document that governs the AI’s behavior and openly entertains the moral implications of its potential consciousness.

Anthropic embedded popular workplace apps like Slack, Figma, and Asana directly inside Claude, allowing users to create designs and manage projects within the chat interface.

Kimi launched K2.5, a new open-weights model that surpasses leading models like Opus 4.5 and GPT-5.2 on key benchmarks and features powerful multi-modal vision capabilities.