Research
Compare the strongest arguments for and against a four-day workweek for a 200-person software company.
Multiple models surface different assumptions about productivity, recruiting, customer coverage, and operational risk.
Examples
Prompts that work well in AI Debate
AI Debate is useful when a question has competing priorities, ambiguous evidence, or room for reasonable disagreement. These examples are designed to produce answers that can be compared, challenged, and synthesized.
Product
Should a small SaaS team build an enterprise SSO feature now or improve onboarding first?
The panel can separate strategic upside from implementation cost and call out missing data before a roadmap decision.
Writing
Rewrite this launch announcement for technical founders, then critique the clarity and credibility of the message.
One review pass exposes vague claims, overstatement, and audience mismatch before the final verdict.
Learning
Explain transformer attention to a high-school student, then identify where the analogy breaks down.
Consensus helps you keep the simple explanation while preserving the caveats a single answer might skip.
Planning
Create a realistic 30-day plan to validate a marketplace idea without paid ads.
Different models tend to emphasize research, outreach, analytics, and operational steps in different proportions.
Tradeoffs
Evaluate SQLite, Postgres, and DynamoDB for a local-first notes app that may later sync across teams.
The agreement map makes it easier to see which constraints drive the recommendation.
What makes a good AI Debate prompt?
- Ask for a decision, comparison, critique, plan, or explanation with caveats.
- Include relevant constraints, audience, risk tolerance, and success criteria.
- Avoid private data, secrets, regulated information, or anything you cannot share with third-party AI providers.