I think Pickaxes are superior because we have access to curated knowledge bases, but it would be great if we could tweak inference time + intermediate reasoning tokens as well to outright beat what Deepseek/o1/o3 are capable of.
EDIT: Yes I know we have access to o1/o3-mini/Deepseek as models, but I guess I’m asking if Pickaxes can/will have more agentic behaviour where they can be triggered remotely and/or perform tasks without attention/supervision in future.