The optimization loop uses Bayesian acquisition policies—specifically Thompson Sampling, Upper Confidence Bound (UCB), and Epsilon-Greedy—to select which snippet or page-edit hypothesis to test next.
flowchart LR
A[Seeded Hypotheses] --> B{Acquisition Strategy}
B -->|Probability Matching| C[Thompson Sampling]
B -->|Optimism in Uncertainty| D[UCB]
B -->|Exploit + Random Exploration| E[Epsilon-Greedy]
C --> F[Selected Hypothesis]
D --> F
E --> F
F --> G[Ideator Craft & Rank Test]