How MindParallax works
Independent views, bounded follow-ups.
MindParallax sends each question to the selected lenses independently. A follow-up includes only the user’s prior questions and that lens’s own accepted answers; other lenses and the Speaker remain outside its context.
Independent fan-out
The system gathers every response separately so a strong answer cannot steer the rest of the group. The lens set stays locked during a thread, and every turn produces a fresh comparison of views that may agree, conflict, or expose different assumptions.
Speaker constraints
After member responses are complete, the Speaker compares their stated reasoning. It can summarize common ground and tensions, but it does not vote, select a winner, recommend an action, or add a hidden consensus.
Demo, live use, and failures
Demo mode uses deterministic placeholders to show the workflow. Live mode uses the configured model provider. If an individual member or the Speaker fails, successfully completed member work remains visible; a failed component is identified rather than silently replaced.
No verdict by design
MindParallax is decision support, not delegated decision-making. It helps reveal considerations; it does not certify a conclusion or substitute for domain expertise, professional advice, or your own judgment.