#35 Remotion, Claude Code Skills, Cloudflare + Astro, Ads in ChatGPT, Apple Siri Chatbot

This week, we explore a range of practical and impactful developments in the AI, design, and development space. We start with a hands-on project using the Blender MCP (Model-Code-Prompt) to turn a simple, hand-drawn sketch of a room into a fully realized 3D model. The AI was able to interpret contextual details from the sketch, such as the meaning of a number indicating window height and identifying IKEA furniture models by name alone, demonstrating a sophisticated level of understanding. While the process was impressive, it also highlighted challenges, particularly in generating complex objects like a bed from product images, which required a more structured JSON-based prompt to achieve the desired result.

From 3D modeling, we shift to motion design with Remotion, a powerful tool for creating videos from React code. The conversation focuses on the new Remotion "skill" available for AI coding assistants like Claude Code. This skill provides the AI with best practices and guidance, enabling it to animate static Figma designs into dynamic videos with simple, high-level prompts. This is part of a broader trend toward AI "skills," further exemplified by Vercel's new 'skills.sh'—a directory and package manager for these skills. This allows developers to equip their AI agents with specialized knowledge for various frameworks and tools, leading to more consistent and higher-quality code generation.

In industry news, Cloudflare has acquired Astro, a modern web framework known for its performance on content-heavy sites, positioning it as a strong competitor to Next.js. Given that Webflow utilizes Astro for its platform and maintains a close partnership with Cloudflare, this acquisition sparks speculation about a potential future acquisition of Webflow, which would further consolidate their interconnected ecosystem.

A major topic of discussion is the recent announcement that ads are coming to ChatGPT. While OpenAI has released principles assuring users that ad content will be separate from AI responses and that user data will be protected, the move has been met with skepticism. The introduction of ads, even in a new, cheaper "Go" tier, raises concerns about user privacy and the potential for the AI's utility to be compromised by commercial interests. The hosts also touch on the nuances of ad relevance versus personalization and how users might react to this change.

Looking ahead, rumors suggest Apple is developing a new Siri-powered chatbot for iOS 27, which will be built on Google's Gemini model. In typical Apple fashion, the chatbot is expected to be deeply integrated into the operating system rather than being a standalone app. This focus on seamless, context-aware integration could make it a highly practical and useful tool for daily tasks, potentially giving it an edge in user experience even if its underlying model isn't the most powerful on the market.

Finally, the episode closes with a critical discussion on AI and censorship. One of the hosts shares a personal experience of being blocked by ChatGPT while trying to write about political protests, a situation that felt like an infringement on personal thought and expression. This leads to a broader conversation about the moral responsibilities of AI platforms and the challenge of implementing a universal ethical framework. In contrast to OpenAI's opaque policies, Anthropic's public "Constitution" for its AI model, Claude, is highlighted as a more transparent approach, giving users clear insight into the values and principles that guide the AI's behavior.