[P] have a couple technical questions for my LLM router. [P]

Explain Like I'm Normal

A developer is building a code-specific LLM router that uses Bloom’s Taxonomy to categorize prompt complexity instead of using expensive model-based classification. By identifying whether a query requires 'remembering' or 'creating,' the system can route tasks to cheaper or more powerful models accordingly. This heuristic approach seeks to optimize token economics for massive codebases where routing overhead usually kills margins.