https://www.youtube.com/watch?v=8JuWdXrCmWg
In short: a new trick we can use for some algorithms, so that we use less memory at cost of longer compute time.
Link to paper:
ECCC - TR25-017