3 previous messages

→ LLM even does napkin math for the instance and calculates running cost for
the Job
→ these Jobs will be re-ran you can use Buckets
with hf-mount that allows fast read-write