I gave myself an AI advisory board — three models argue, I decide

Short summary

The author runs three independent LLM models (Gemini analyst, Opus architect, Codex builder) in parallel to stress-test decisions before committing. Value comes from disagreement: each model reveals different failure modes that a single model would miss. Rather than trusting one confident voice, this approach lets humans maintain decision authority while leveraging AI for multi-angle risk assessment.

• •Run three independent models with distinct roles to stress-test decisions from different angles
• •Value comes from disagreement—conflicting perspectives reveal failure modes a single model would miss
• •Humans synthesize and decide; AI provides risk assessment without outsourcing judgment

Generated with AI, which can make mistakes.