https://www.youtube.com/watch?v=KIlI46HdqKg
OK, so I have finished watching the video and I will try to describe what is in it. The author has done a deep dive into the votes for Philadelphia on Nov 4 using the data on the NY Times website.
The author notices suspicious patterns. The votes are coming in at exact fixed ratios across multiple precincts. For example exactly 1 Trump vote per exactly 6 Biden votes. These "fixed ratio batches" are coming in for multiple precincts in discrete time segments. So, for say 2 hours, votes into a certain precinct are coming in at exactly at this 6:1 ratio. Then moments later the computer reporting votes in a different precinct at exactly this 6:1 ratio. It's not "averaging out at 6 to 1" like you might expect from an actual count, but it's coming in precisely fixed, exactly 6 Biden votes to 1 Trump vote. Again and again.
The author has identified an number of other ratios that were used to stuff the ballot boxes until the desired total was reached. At the 36 minute mark the author describes the scheme he's observed.
So it's a partial deconstruction of the Dominion vote stuffing or vote switching algorithm. Towards the end of the video, if you want to fast-forward, the author draws a helpful diagram that indicates how the scheme appears to work.
EDIT: The author of the video made this helpful summary graph https://ibb.co/h1x3Xds and he actually posts here https://thedonald.win/p/11QRkGgVT6/smoking-gun-i-discovered-dominio/
EDIT2: Let's get more eyes on this and see if we can find other cities where this type of fraud happened and see if we can make it into an easier-to-understand summary.
If you look at the data for Philadelphia and the 4th alone, then there are only 4 times when the system is actually updated with new data. Inbetween the data stays the same for every polling location. Therefore all these rows inbetween can actually be deleted.
But if you have only 4 updates, then it is logical that the ratios "transfer" during these updates.
Pls go and verify this yourself. Heres the amount of polling places reporting actual updates per each timestamp:
9:39:44 -> 1685
15:13:33 -> 1695
20:10:16 -> 189
20:47:22 ->1677
Again, for all other timestamps there was no actual change for no polling location.