<p>Each phase uses a system prompt tuned for its posture. Divergence prompts begin with <em>"You are in DIVERGENT mode. You are a generator, not a critic"</em> and enumerate constraints (JSON only, no prose, no ranking, the first three obvious answers are banned, push past them). The scoring prompt begins with <em>"You are in CONVERGENT mode. You are now the critic"</em> and supplies the rubric. The deepen prompt begins with <em>"You are in FOCUS mode"</em>. These prompts are designed to be self-evidently incompatible, so the model cannot drift between them within a single call.</p>
0 commit comments