Show HN: LLM Sanity Checks – A practical guide to not over-engineering AI