FFmpeg

Claude Fable, Claude Tag, and Anthropic's Culture — Cat Wu & Thariq Shihipar ft Simon Willison

Jul 15, 2026 · 51:30

This episode features Anthropic's Cat Wu and Thariq Shihipar discussing Claude Code, Claude Tag, and Claude Fable, arguing that these tools have shifted engineering from slow spec-driven processes to rapid, ambitious building. Thariq notes that with each model generation, delegation increased, and Claude Fable now enables one-shot features. Cat says engineers now need product taste over execution, as timelines shrink from six months to a week. Claude Tag, a proactive multiplayer agent, lands 65% of product engineering PRs internally by monitoring channels and remembering preferences. The team reduced Claude Code's system prompt by 80% for frontier models by removing examples and hard constraints, relying on model judgment. Auto mode, used internally since January, mitigates prompt injection through thousands of evals and Sonnet classifiers. Cultural hacks include default-public channels and a 'don't negotiate against ourselves' mindset, leading to ambitious builds like Thariq's Claude-powered video editing and a Street Fighter game.

This video was edited with AI agent. But how?

Feb 22, 2025 · 5:00

Muhteşem from Re-Skill presents the world's first open-source video editing agent, built in collaboration with Diffusion Studio, which uses LLM-generated code to edit videos via a browser-based UI. The agent starts a Playwright browser session and connects to operator.diffusion.studio, a web app built with Diffusion Studio Core that renders video directly in the browser using WebCodecs. Three tools drive the process: VideoEditingTool generates and runs JavaScript code based on user prompts, DocsSearchTool uses RAG to pull relevant information from llms.txt, and VisualFeedbackTool samples one frame per second and decides whether to proceed with rendering or refine further. File transfers between the Python agent and browser happen via Chrome DevTools Protocol, and for scalability, the agent can connect to a remote GPU-accelerated browser through WebSocket. The project emerged from Re-Skill's need for automatic video editing for personalized learning, where they found FFmpeg too limiting and turned to Diffusion Studio Core for its intuitive API and reliable in-browser rendering.