A couple of days ago, Cursor dropped Cursor Composer 2, and I wanted to test it in a way that actually matters: can it build something real that people would use (and pay for), or is it just another hype tool? I walk through the basics—Composer 2 being cost-efficient and fast (with the chart showing it pushing ~200 tokens/sec on “fast”)—and then I do the live experiment: no manual coding, just Cursor Composer 2 and me. If it breaks, you see it. For the build, I keep it simple on purpose: a clean web app where you paste long text and get a short summary. I start with a plain-English prompt (“build a clean web app…”) and let the agent plan, scaffold, and write the React app. We hit real issues (localhost not loading, 500 errors, port problems), and I show how I prompt the AI to debug and fix it until it runs properly. Then I push it further: the first UI is messy, so I prompt it to make the UI modern and visually good—and it actually upgrades the design a lot. By the end, the summarizer works, clear/reset works, and we even get stats like original characters vs summarized characters. My takeaway: Cursor Composer 2 can build a simple SaaS-style app in minutes with a few prompts, and that’s enough to validate whether you should try bigger, more complex builds next.