 So what did we find? Big success at fooling the discriminator and complete failure at making realistic images at least to humanize. So now let's iterate that, like what we'll now do is we'll take the discriminator generator that we just trained for the Lawn discriminator that will hopefully draw better images And then we're gonna freeze this and now we will lawn a better discriminator. So we lawn the discriminator generator that we just trained for the Lawn discriminator that will hopefully draw better images at least to humanize.