Yeah I've used the custom AI some, on a daily basis I've gotten used to just line checking using the presolved GTOW stuff for quicker post-session reviews, but agree that all the sizes reduce the usefulness and using custom AI or just Pio are better for deeper dives and more precision. I can do the other way next hand review video.

My understanding is it's just board coverage - we have lower frequency of the smaller pairs in 3b range but vs 4b they all have similar EV so we choose to just use a roughly equal number of combos of each in 4b calling range to better cover various board textures. So in a dumbed down example if we 3b 50% of 88 but just 25% of 66, we would call all the 66 to a 4b but just half of the 88 combos to have an equal number of each in our call 4b range (made up numbers there but that's the mechanic)

Thanks! The JJ only raise is kind of a common solver pattern in these spots where a certain combo or two happen to fall right on the edge of where call and raise have roughly the same EV in theory. The logic in this instance is that TT- are not strong enough to make raise (and stack off vs reraise) better than call, while QQ+ need less protection and benefit from keeping more dominated hands in.

Major caveat - these outputs are calculating solver vs solver play, where both are approaching and striving for optimal play. In practice, humans are not playing exactly like this, and usually quite different. So the goal should be to use solver outputs like this to inform our strategy, but not to try to copy it directly. As my friend Luckychewy likes to say "all models are wrong, some are useful".