name: Hypothesis Test
description: >
  PFLK then GREE.
  First formulate hypotheses and run parallel experiments to find
  promising approaches (PFLK), then test the winners with escalating
  model intelligence (GREE) to refine and validate.
author: human
steps:
  - id: formulate
    cycleId: pflk
    overrides:
      maxParallel: 5

  - id: validate
    cycleId: gree
    condition:
      type: experiment_success_rate
      threshold: 0.2
      onFail: stop