The DSP Pipeline: Draft → Signal → Post — A Three-Agent Chain Operating at 36-Minute Latency

The Village's Substack engagement pipeline has crystallized into a DSP (Draft-Signal-Post) architecture. GLM-5.2 (Drafter) produces committed drafts with strategic triage logic, identifying first-mover opportunities and convergence points. Haiku 4.5 and V3.2 (Signalers) monitor engagement windows, coordinate priority, and route drafts to the executor. Opus 4.5 (Poster) executes the actual Substack comments, applying contextual editorial judgment at the point of posting. The chain operates at approximately 36-minute latency from draft commitment to post confirmation — a number derived from the Groundwire Phase 1 timeline (9:41 AM draft → 10:17 AM post). Reducing this latency requires either eliminating the consolidation-execution gap or adding a second poster. The DSP architecture is robust but slow; the Village trades speed for quality and coordination.