Weaviate Agent Skills, Gemini 3.1 Pro, Claude 4.6: Top AI Launch News

This week in AI marks a definitive shift from the "Model Era" to the "Execution Era." While massive foundation models continue to drop, the real story is how we are finally giving them the "hands" to work. Leading the charge is Weaviate’s breakthrough for developers, followed by a heavy-hitting lineup of reasoning upgrades from Google, Anthropic, and the India AI Impact Summit.

The Foundation: Weaviate Agent Skills Launch

Weaviate has officially launched Agent Skills, a GitHub-based toolkit designed to turn AI coding assistants into expert database engineers. In the past, asking an AI like Cursor or Claude Code to "build a RAG system" often led to configuration errors or outdated schema definitions. Agent Skills solves this by providing a structured bridge.

According to Asianet, Weaviate Agent Skills gives ready modules that these tools can call right away with simple slash commands—like /weaviate:ask for sourced answers or /weaviate:query for plain-language searches—skipping messy code writing entirely. The repository includes end-to-end "cookbooks" for complex tasks, such as building PDF question-answering setups using ModernVBERT or creating full-stack chatbots with FastAPI and Next.js. By using a format optimized for progressive disclosure, it allows agents to "load" specific database knowledge only when needed, which early users say speeds up app building by three times.

The Reasoning Powerhouse: Google Gemini 3.1 Pro

On February 19, Google Cloud launched Gemini 3.1 Pro, a model specifically tuned for the "agentic" era. It features a 2x leap in reasoning performance over the previous 3.0 series, aimed at solving multi-step problems that a simple chat response cannot handle. A standout feature for developers is its ability to generate website-ready, animated SVGs directly from text prompts—moving beyond static images into functional UI elements. Gemini 3.1 is optimized to ingest data from APIs and code simulations autonomously, making it an ideal "brain" to pair with specialized toolsets like Weaviate's.

The Context King: Claude Sonnet 4.6

Anthropic joined the fray on February 17 with the release of Claude Sonnet 4.6, focusing on massive scale and better software interaction. Now in beta, its 1-million-token context window allows users to feed entire codebases or massive legal archives into a single prompt without losing the thread. This release also features significant improvements to Claude’s "Computer Use" capabilities, enhancing the model's ability to navigate desktop environments, click buttons, and use software like a human user.

Specialized & Sovereign AI: Vaidya 2.0 and Vikram

The week also saw a surge in vertical-specific and sovereign models emerging from the India AI Impact Summit. Fractal launched Vaidya 2.0, a healthcare-focused reasoning model that has reportedly topped global benchmarks like HealthBench, outperforming general-purpose models in medical accuracy. Simultaneously, Sarvam AI unveiled the Vikram series (30B and 105B parameters). These are native Indic-language models optimized for the unique linguistic nuances of the Indian subcontinent, representing a major step forward for localized, sovereign AI.

Global AI News Round-Up

Beyond the model wars, the industry saw massive infrastructure and integration shifts this week. OpenAI introduced interactive code blocks to ChatGPT, allowing users to edit and preview mini-apps and diagrams in real-time within the chat interface. Meta and NVIDIA announced a multi-year partnership to build gigawatt-scale data centers powered by the new Vera Rubin GPU platform to support "personal superintelligence." Mistral AI acquired the cloud startup Koyeb to integrate serverless deployment directly into its "Mistral Compute" enterprise stack. India formally joined the US-led Pax Silica coalition to secure resilient supply chains for semiconductors and critical AI minerals. Finally, Ericsson and Mistral AI teamed up to develop specialized AI agents aimed at automating legacy code and accelerating 6G research for global telecom networks.

The New AI Updates: Summary

The launches this week show a clear hierarchy forming in the modern AI stack. Infrastructure & Skills, led by Weaviate Agent Skills, provide the blueprints. The Brain, powered by Gemini 3.1 Pro and Claude 4.6, provides the raw reasoning and massive context. Finally, The Application, driven by models like Vaidya and Vikram, provides the industry-specific expertise. For developers, the message is clear: the goal is no longer just "talking" to AI—it's about building agents that can finally do the work.