Ideas for Sources Outside LLM for Reflection Pattern

In Agentic AI module 2 Andrew explains that going outside the LLM for reflection is powerful and assists the LLM in refining the output. The use case given was to run code. Curious what other scenarios people are using this pattern for and what sources they are using that do not relate to code?