Write your output to:
  {report_path}

Output format: One learning per line, starting with "- "
Be concise (max 20 words per learning).
Only include learnings that generalize beyond these specific tasks.
Do not include task-specific details that won't apply to future work.
