[emodendroket]

That's interesting on Deepseek. But I think as long as the models are still making noticeable gains with each iteration it's hard to say "good enough."