compound-learning

A self-improving knowledge accumulation workflow coordinates Plan, Execute, Assess, and Compound loops to crystallize knowledge. It stores documented solutions in a knowledge base, enforcing YAML frontmatter schemas to drive continuous improvement across repeated tasks.

The best way to codify documented solutions uses YAML frontmatter schemas and a docs/solutions directory for retrieval and reuse. This knowledge codification captures learnings from task assessment, making accumulated knowledge available for future writing workflows.

Yes, you can use modular agents and parallel workflows for multi-domain learning and design tasks. The compound-learning approach supports coordinating these parallel workflows across domains where task quality matters and patterns emerge from repeated work.

The Plan Execute Assess Compound loop improves future work quality by enforcing a structured cycle. It assesses task execution, compiles learnings into a knowledge base, and applies accumulated knowledge to enhance subsequent planning and execution.

Yes, you need a frontmatter-driven knowledge schema format to start a compound-learning cycle. The workflow requires YAML frontmatter schemas to codify knowledge and output documented solutions under docs/solutions for effective retrieval.

You should not use structured knowledge crystallization for one-off tasks lacking repeatable patterns. This approach applies across domains where patterns emerge from repeated work, making it less suitable for isolated, non-recurring problem solving.