Hadn't they spent a year hardening curl with various AI before they tried Mythos?

Yes. The original curl post didn't say anything like "mythos sucks" but rather "it's only a minor improvement in comparison to already widely used models".

Yes, and Firefox had not.

Which I think points at Mythos not being some big jump in capability finding things earlier LLMs didn't, it seems to mostly come down to massively increased compute budget and they finally catching up in context sizes.