I saw a big hit to Claude’s intelligence w/ the 1M context window model and the change to adaptive reasoning (github issue linked elsewhere in this thread). I’

I saw a big hit to Claude’s intelligence w/ the 1M context window model and the change to adaptive reasoning (github issue linked elsewhere in this thread). I’m pretty much using 90% Codex now, although since Claude is consistently faster at answering quick questions, I still keep it open for that and for code-reviewing codex/human work before commit.