Executive Summary
Paper XVIII showed that priming selection closes 48.9% of the coordination gap through processing style alignment. This paper dissects the structural features of effective priming sequences to identify what makes a coordination sequence work. The goal is to move from empirical priming selection toward principled protocol design.
The key discoveries overturn several intuitive assumptions about coordination. First, effective priming is remarkably compact: just 15 tokens capture 98.8% of the full benefit, meaning that coordination is about triggering a mode switch rather than transferring detailed information. Second, natural conversational phrasing outperforms rigid ritual formatting by 0.48 nats, suggesting the model processes natural language more effectively than structured commands. Third, repetition is actively harmful, degrading performance by +0.07 nats per repeat — the model treats repeated instructions as noise rather than emphasis.
The optimal protocol that emerges is simple: reset then prime. A reset instruction clears the receiver's prior context, followed by a minimal domain-appropriate priming sequence. This achieves cross-entropy of 0.733, beating expert-designed coordination baselines by 39%. Forced domain naming is counterproductive (+0.568 nats worse than natural vocabulary), reinforcing the finding from Paper XVIII that effective coordination operates through implicit processing style alignment rather than explicit domain labeling.
Key Findings
- Extreme compactness: 15 tokens capture 98.8% of coordination benefit — coordination is a mode switch, not information transfer
- Natural beats rigid: Natural conversation outperforms rigid ritual formatting by 0.48 nats
- Repetition degrades: Each repetition adds +0.07 nats — the model treats repeated instructions as noise
- Forced naming harmful: Explicit domain naming worse than natural vocabulary by +0.568 nats
- Optimal protocol: Reset instruction followed by minimal priming achieves CE 0.733, beating expert baseline by 39%
Key References
- McEntire (2026) — Ensemble Gravity: priming selection dominates coordination (Paper XVIII)
- McEntire (2026) — Shepherd Agents: adaptive priming is harmful (Paper XX)
- McEntire (2026) — Context Fences: mode switching mechanism (Paper XXII)
- McEntire (2026) — The Coordination Problem Is Interference: synthesis (Paper XXI)