> Over time we also independently adopted a habit to isolate LLM output > into dedicated commits, that proved very useful during post-mortem and > unexpected-costs analysis (bugs, security incidents...). I'm curious, does this mean that raw LLM output and the corrected version of the LLM output are in separate commits?