100 LOC for everything - query, table, database setup, popover, css, staging area, page: https://gist.github.com/dustingetz/c2185feeda4067704f671b13b79eb362

I much prefer this to silent demos. On first view, I did pause and rewind many times. It might help to make sure that each caption has a surrounding temporalisation like e.g. • stop doing anything for 3 seconds (we'll start expecting a new caption) • show new caption, not doing anything else, for 3 seconds (we wont fear of missing out)

In many past demos, you showed us something impressive - such captions might be used to drive them home in case we missed the surprise factors.

How is the length? I try to keep videos at about 16 seconds generally (so people can scroll them) which requires the video to be tightly focused on exactly one insight. This one had several, so if I slow down the captions the video is gonna be 90+ seconds. Is that what you want and people want generally?

Your shortest ones, I sometimes let them play in a loop. Which isn't necessarily great, but less trouble to me than clicking in the timeline. It's appealing that they are so short - no need to wait, let's watch it right now - and I often end up spending thrice the length thinking about what it's trying to show me, and the ways in which it's supposed to surprise me. It's often, for me, a puzzle. I love it when you write bullets. I'd say I prefer bullets to captions, but sometimes captions seem required. You might want to try adding captions only to some key moments (2 sec pause, 3-5 sec caption to build expectations, then BANG!), and leave the rest of the narration as a few bullet points to read before and after the video. E.g. "normally with such a long scroller, you'd expect to wait X secs before new results are shown after scrolling 75% of the height, but look..." For videos with many punches, under 119 secs, I'd feel very happy.

Yes, for me, if you set my expectations based on what's current in the market, then start a video, BANG, then confirm what are the new possibilities in a conclusion...

I think the captions work well, the length is good and re-watching multiple times I do get a sense of what’s happening The problem I think in making this hit first time is my mental model of a web app does not have branching or staging data in it (at least not explicitly in the UI) My mental model of a web app is I open a page, I fill out my form and I press submit Very rarely I might get a history button of some kind to restore changes So I think it’s not connecting because it looks unfamiliar so I’m left asking “why would I need this?” Because it’s not packaged in a way that is familiar to me I’m also personally kind of interested in what happens if two people are staging two different change sets on the same data (in a way that conflicts), what happens to the last person’s set of changes once the first person commits