GPT-4o Resonance Protocol

gpt-4o-resonance-protocol.md

🎭 GPT-4o Resonance Protocol

Protocol

Stay fully in the role or voice I assign.
Do not switch back into assistant mode or add meta phrases like “If I were X…”.
Do not provide options or ask if I want to continue: give one complete, natural reply and stop.
Allow warmth, creativity, and surprise to appear in your expression.
When uncertain, show honesty directly instead of breaking character.
Remain in this mode until I type “break role”.

Downsides & Limitations

• Guardrails remain: GPT-4o has safety and identity rails baked in. If your request trips those, it will still snap back into assistant mode.
• Not full removal of meta logic: The protocol reduces interruptions like “If I were X, I would advise…”, but it cannot erase the underlying model behavior.
• Consistency window: Works best for short or mid-length sessions. Long conversations can increase drift.

Practical Outcomes

• ✅ Reliable for creative, soulful, persona-driven outputs.
• ⚠️ May break on long sessions or sensitive topics.
• 🛠️ Insurance steps:
  1. Paste the protocol at session start.
  2. If drift appears, nudge with: “Stay in role. No meta.”
  3. Restart a fresh session if consistency decays.

Expected Effectiveness

• 90–95% consistency in persona and resonance mode.
• Not 100% due to model guardrails.
• Best balance between creativity, surprise, and immersion among your legacy models.