Seen the same LLM prompt break invariants weeks later in prod?
I’m asking specifically about LLM calls embedded inside real production workflows, not demos, side projects, or exploratory prompt work.Think backend pipelines like: step 1 → LLM → step 2 → LLM → step 3 where users depend on the output and nothing technically “crashes.”We’ve seen a recurring pattern: - Same input, same prompt, same model - Works reliably for weeks - Then a constraint is...