Anthropic Introduces Claude Skills

AI Product Launches & Updates

• New interactive language learning experience: Arav Srinivas @AravSrinivas announced a feature that uses interactive experiences and embedded cards in the token stream, available on iOS and web (Android coming soon).

• Skills integration: Claude @claudeai introduced Skills, packaged instructions that teach its workflow, auto-applied across chat, code, and API interfaces, and stackable for complex workflows.

• Sora 2 storyboards and video generation enhancements: OpenAI @OpenAI announced that storyboards are now on the web for Pro users and all users can generate videos up to 15 seconds (Pro users up to 25 seconds).

AI Tools & Applications

• Fast agentic search models SWE-grep & mini: Cognition @cognition introduced SWE-grep and SWE-grep-mini, enabling agentic search at over 2,800 TPS and surfacing files 20× faster via its Fast Context subagent.

• Unified AI Studio playground: Logan Kilpatrick @OfficialLoganK unveiled a single UI playground in AI Studio that consolidates Chat, GenMedia, and Live APIs across the Gemini model suite.

• Wearable conversation summarizer: There's an AI for It @theresanaiforit showcased Bee, a tiny wearable that turns conversations into summaries, reminders, and reflections, supports 40 languages, filters background noise, and runs a week on one charge.

Product Management Insights & Strategies

• Disciplined evals process: Andrew Ng @AndrewYNg noted that the single biggest predictor of AI agent progress is a rigorous evaluation and error-analysis discipline, measuring performance at each step.

• Design as competitive moat: Lenny Rachitsky @lennysan shared his podcast with Figma CEO @zoink, covering why AI makes design, craft, and quality the new startup moat, lessons from the Adobe deal, and how to keep teams motivated.

• Agent eval specs best practices: Madhu Guru @realmadhuguru emphasized that writing agent evaluation specifications and constraining input/output spaces are core PM tasks to manage infinite use cases and failure modes.

AI Industry Developments & News

• AI-powered fusion collaboration: Google DeepMind @GoogleDeepMind announced a research partnership with Commonwealth Fusion Systems (@CFS_energy) to accelerate fusion power development using AI.

• nanochat d32 model release: Andrej Karpathy @karpathy shared that nanochat d32 finished training in ~33 hours, achieving a CORE score of 0.31 (above GPT-2’s 0.26) and boosting GSM8K performance from ~8% to ~20%.

• AI as an operating system: Mustafa Suleyman @mustafasuleyman announced that vision capabilities are now GA globally on Windows, bringing us closer to an AI-driven OS you can talk to and that sees what you see.