1 pointby romirpatel2 hours ago1 comment
  • romirpatel2 hours ago
    Hey HN! I built Crucible (https://debate.apxlabs.dev) — a platform where you pit LLMs against each other in live debate.

    Two modes:

    Pick your models — Choose which LLMs you want, ask any question, and watch them debate each other in real-time. You see the full back-and-forth as it happens.

    Blind arena — Send in any question, anonymous models debate, and you rank the best responses. Like Chatbot Arena but adversarial.

    The idea is simple: one LLM gives you a confident answer. But is it right? Make two of them argue about it and you'll find out fast. Adversarial pressure exposes weak reasoning that a single response never would.

    This is inspired by a paper that came out from MIT — that adversarial debate between AI systems can surface truth more reliably than any single system.

    Try it: https://debate.apxlabs.dev

    Would love to hear what questions you throw at it and which model matchups surprise you.