I kept hearing about Cursor, this “AI IDE” that’s supposed to be way better than a normal Copilot setup because it actually understands more context across your whole codebase. So I decided to put it to a real test: could it help me ship an MVP for a startup idea I’ve wanted for a while—an app that generates SRT subtitles from a video upload so I can just import them into DaVinci Resolve and stop wasting time matching timings by hand. The first hiccup was cost. The obvious route was calling OpenAI’s Whisper API, but I didn’t want to pay per request just to subtitle my own content (at that point I might as well subscribe to an existing tool). Cursor came in clutch by pointing me to whisper.cpp, a lightweight C++ implementation that runs locally and offline. I wired up a simple backend that converts video to WAV with ffmpeg, calls the whisper CLI with the English model, and outputs an SRT. On the frontend, I kept it super MVP—plain JavaScript, basic folders (server/client), and just enough UI to upload and get the file back. Cursor wrote a lot of the code fast, but I still needed real dev experience for setup work like installing dependencies, managing the submodule, and running everything. End result: it worked, and I subtitled the entire video using the tool I built.