Code executes, the LLM judges

Build-along · ~10 min at your own keyboard

Models are probabilistic; your invoice math shouldn't be. The split: scripts handle counting, sorting, fetching, formatting (deterministic, ~free, 100,000x faster); the model handles judgment — tone, prioritization, what-matters. Reserve LLM calls for decisions.

Do this

0/3

Share what happened

Paste your two-column split and the twice-run consistency result.