In this video I walk through how I built a fully automated video production pipeline using Claude Code, HeyGen, and ElevenLabs. You'll see how Avatar V creates realistic AI clones, how to get voice cloning that actually sounds like you, and how Claude Code orchestrates the entire process so you can go from raw script to finished video overnight. I also cover the real costs, the limitations, and the objections you're probably already thinking about.

80 nodes in n8n. Now zero. A few months ago I built a system to handle collaboration and sponsor emails automatically. It worked. But it was 80 nodes, hard to debug, and one wrong step from breaking. The pain: collab emails come at all hours. Miss one and you lose a deal while you were sleeping. Last week Claude Code rebuilt the entire thing in Python. Deployed on Modal. Now it runs every 15 minutes, reads my inbox, classifies every email by intent, and sends me a Telegram message with approve/reject buttons. During my active hours — I review and approve. During off hours — it follows base guidelines I set and replies on its own. Lead stays warm. Deal doesn't go cold. Zero n8n nodes. Zero missed opportunities. This is what building with Claude Code actually looks like in practice.

What you're seeing right now is not the official Apple website. I built this in 30 minutes with Claude Code. 192 frames. Scroll-driven canvas animation. A MacBook Neo gliding through the screen as you scroll, perfectly synced to every pixel of movement. Smooth transitions, massive typography, the whole thing. Websites like this used to be a nightmare to build. Not years ago. Last year. Setting up GSAP, canvas frame rendering, Lenis smooth scroll, getting the product video to sync frame by frame with scroll progress... just the boilerplate alone took half a day. Making it actually look premium was a different battle. I decided to solve this once. I built a skill inside Claude Code that knows the entire stack, the design rules, the animation logic, everything. You give it a product and a video. It does the rest. The first output genuinely shocked me. Let me know what do you think in the comments.

Boris Cherny is the creator of Claude Code by the way :)

I just built an AI chatbot that speaks Telugu. Not just types in Telugu. It listens to you, thinks, and speaks back to you in Telugu. Out loud. In real time. I recorded a 2-minute demo walking through the whole thing and honestly, watching it work still surprises me. Here is what happens under the hood. You open the chatbot, tap the mic, and speak in whichever language you choose. English, Hindi, Telugu, Tamil, Kannada and more. The bot understands you, searches its knowledge base for the right answer, and speaks back to you in that same language with live word-synced subtitles on screen. For English, it uses ElevenLabs to create a natural, human-sounding voice. For Indian languages it uses Sarvam, which is purpose-built for Indian languages. And behind the scenes, n8n orchestrates the whole pipeline, with OpenAI powering the intelligence. Here is why I built this. Most AI chatbots are designed for English-speaking users in Western markets. But India has 1.4 billion people, and the vast majority of them are far more comfortable in their mother tongue. A chatbot that only works in English is leaving most of your potential customers behind before the conversation even starts. This is what I built for SSK AI Agency, a fully multilingual voice AI assistant that knows your business, speaks your customer's language, and is live on a website right now. If you are running a business and want something like this, drop me a message. I am taking on a few clients this month and would love to show you a live demo. What language would you want your chatbot to speak first?