[bongodongobob]

I played it with 3.5 and it was great. This isn't something o1 just picked up on.

[suveen_ellawela]

yep, agree. One big part of the experiment was to see how well it does the reasoning by asking it to output the reasoning.