@decent-artlu
it is becoming likely that I will vibe software three times or more:
1. in openclaw-inspired state of mind, where agents are unbounded by possibilities AND the threshold for getting a POC/MLP is flat/cheap
2. in controlled research environment, with evals + testing to refine what stage 1 produced
3. in tighter production, where we actually rebuild the logic flows (SPEC.md) rather than remove pieces of scaffolding from stage 2
Software eng used to have `dev` and `prod`, maybe with a `staging` that was really a substitute or fallback for catching mistakes earlier or later.
it really should be different frameworks for `prototype` -> `research` -> `deploy`
e.g., OpenClaw -> Mastra -> n8n