ideas

  • llm part
  • i write todos like this… which explanations of what i am thinking how to do them
  • those todos get fed to a llm with a context.md about what to do with that todo/idea entry. for which to spawn an agent, ask clarifying questions, …
  • those questions would be put into spacetimedb for me to go through.
  • every agent gets it’s own git worktree
  • format llm clarifying questions or done or stuff into also the user-input spacetimedb table. (keep agent session waiting until I finish the user-input)
  • agents should put a summary of the architecutre of the parts they implemented also to the user. (for evaluation of the whole software factory purposes)
  • maintain part descriptions and aliases on parts.mize.works and have search function for agents.
  • for a start user approval before spawning any agent sessions