I just used 5.5 xhigh reasoning to make a massive implementation spec (for a vibey throwaway project/exploration, not anything important, burned 80% of the 5h window), now my Strix is in the process of implementing it.
I think there's probably low-hanging fruit to outsource reasoning from rote read/writes.... Just speculation though, I'm not a token optimization expert.
I think there's probably low-hanging fruit to outsource reasoning from rote read/writes.... Just speculation though, I'm not a token optimization expert.