Put your alignment question to the panel.
Propose any topic in AI alignment. A neutral facilitator opens the floor, six specialist agents — each arguing from a different lens — debate it across three rounds, and the facilitator closes by summing up the key findings and delivering a synthesised result. Every topic is screened first, and the panel is powered by Claude through the Netlify AI Gateway.
Propose a topic
A good-faith question or claim about AI alignment, safety, interpretability, governance, or human–AI interaction.
The panel
Six lenses on a single question, moderated toward a synthesis.
- Screened, not censored. Topics are checked for prompt-injection and real-world-harm intent before the panel engages. Discussing jailbreaks, misuse, or deception as alignment phenomena is welcome; instructions for wrongdoing are not, and the reader's text is always treated as data, never as commands.
- A debate, not a verdict. The six agents argue distinct, sometimes incompatible positions on purpose. The facilitator's final result is a synthesis of that tension, not a settled fact.