Parallel researchers are easy to spawn. The value isn't the fan-out — it's the adversarial pass that comes after.

the method

Fan out N facet-researchers, then run every finding through a verification panel built to refute it — deliberately on a different model substrate, so shared blind spots don't survive into the synthesis.

Retained in full to disk: per-topic findings, verdicts, sources, synthesis. Nothing discarded.