It's done some great things, and when it works, it's solid.
It has failed many times in generations sadly for me. That seems to be happening more now than it did before in basic single input generations.

I'm sticking with it, and it shows a lot of promise. If I tried to do anything with multiple videos, I tended to have problems. But I outside of repeatedly seeing the weird sad face drawing on content failures, I've got no way to even flag the failures for the TubeOnAI team to check into and help fix it.
But early on in testing, the AI agent was able to at least understand the problem I had an took a note for the team to review. I didn't get a follow up, so I'm not even sure if the AI bot got the message to them, but it seemed to at least understand the issue.

Anyways, overall, I've been impressed with this.

Something MAJOR they really need to add is an export text with formatting (on mobile), or a save to PDF function to retain the formatting and avoid going from great output to a generic wall of relatively unformated text.
I think I have to rate it 3/5 due to the extensive generation failures and the inability to export formatted answers from the app interface.

Who would love this the most are the people who are following specific content creators/channels and want to get quick, prescoped summaries and highlights instead of feeling like they have to watch all the videos/content themselves. It can help you determine which ones to invest your time into!