The idea of marking directly on the video frame instead of typing out a paragraph explaining which part looks wrong is the detail that actually solves something real. Anyone who's tried giving feedback on a video knows how hard it is to describe that thing in the corner, no, the other corner in words alone. Circling it and just saying what to change removes that whole problem. Keeping the same conversation going instead of starting a new prompt for every small change is what makes this feel less like operating software and more like working with someone who remembers what you already discussed. That's a different experience than most editing tools, where every adjustment means digging back through a timeline. Templates being a starting point instead of a fixed answer also feels right, since most people don't want to build from nothing, they want a shape to react to.