Stable identifiers
Panel members are pinned to deterministic IDs, which makes reruns, citations, and internal comparison possible.
Antestrata does not generate respondents on demand around a prompt. It samples a persistent synthetic population, pins panel members to stable identifiers, and lets you run that same panel again after the rewrite.
Antestrata builds the world first, then draws the panel from it.
The difference sounds subtle until you use the output. Prompt-first systems can generate a plausible voice on demand. They struggle to give you a stable research object you can revisit.
Antestrata treats the panel member as a persistent behavioural object with biography, contradiction, provenance, and a stable identifier. The run is a draw from that substrate, not an improvisation around your request.
Panel members are pinned to deterministic IDs, which makes reruns, citations, and internal comparison possible.
Reactions are attached to memory and biography. The output shows the graph. The synthesis grammar stays withheld.
A good run names where the sample ran thin. Thin coverage is part of the result, not a thing the UI hides.
A persistent population changes the meaning of a run. The product becomes something you can build evidence with across versions instead of something that performs once.
That is also why the output does not stop at reaction. The method is strongest when it can say: revise this, rerun this, or take this version to real-user testing and start with these profiles.
Sometimes that means revise and rerun. Sometimes it means Antestrata thinks this version is ready for real-user testing, and here are the profiles to recruit.