Multi-Agent
Watch a multi-agent pipeline think.
Ask a question worth debating. Three researcher agents fan out in parallel (optimist, skeptic, pragmatist), an adversarial critic checks their work and returns a verdict, and a synthesizer merges it into a final answer. It's all live: LLM calls on a free model, so your click spends my tokens. Keep it to a question or two.
waiting…
waiting…
waiting…
Fan-out
waiting…
waiting…
waiting…
Critique
Synthesis
How it works
1 · Fan-out
Three researchers run in parallel, each locked to a different lens, so you get genuinely diverse takes instead of one averaged answer.
2 · Adversarial critique
A critic agent hunts for weak claims and contradictions, then returns a PASS/FAIL verdict on whether the research is sound enough to trust.
3 · Synthesis
A synthesizer folds the notes and the critique into one balanced, decisive answer. It's the orchestration pattern I build every day.