Hacker News Reader

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

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