IDEAS.md

ideas that can help

how to improve LLM output quality

— consider each llm output by a separate agent that reasons about output quality (x2 more expensive, but more precise outputs) — split agent into smaller agents