DEJAN classifies optimization hypotheses into a locked two-level taxonomy of 36 distinct factors across 7 classes, modeling each strategic claim with Bayesian posterior beliefs that update based on whether targeted content edits improve model ranking.
flowchart TD
A[Seeder Generates Claim] --> B[Classifier Assigns Factor]
B --> C{7 Taxonomy Classes}
C --> D[Alignment]
C --> E[Substance]
C --> F[Architecture]
C --> G[Style]
C --> H[Framing]
C --> I[Proof]
C --> J[Other]
D & E & F & G & H & I & J --> K[Ideator Rewrites Snippet]
K --> L[Ranker Evaluates vs Competitors]
L --> M[Update Beta Posterior]
M -. Iterative Loop .-> KReferenced by