In this video I walk you through building a ChatGPT-style clone in under ~60 lines using the Vercel AI SDK. The big idea is: Vercel has made AI a priority, and this SDK gives you the “boring plumbing” for free—streaming-first UI, simple adapters, and a super clean API between your Next.js app and OpenAI. I demo the streaming effect (tokens coming in live), then we build the whole thing with Next.js 13 App Router, TailwindCSS, and a single /api/chat endpoint. On the backend, I set up an Edge-compatible OpenAI client (using the fetch-based package) and return a StreamingTextResponse by piping OpenAI’s stream through OpenAIStream. On the frontend, I use the useChat hook from ai/react, loop through messages, and wire up input + handleSubmit—then we style it to look like ChatGPT (including the assistant background + logo). Finally, I deploy to Vercel with the Edge runtime enabled (export const runtime = "edge"), add env vars safely, and ship it. The takeaway: with streaming + Edge, you can build fast AI SaaS experiences with way less code than you’d expect.