Lightweight.
of median provider latency
vs ≈ 4.762 s median provider latency
AI models will change. Providers will change. Reasoning paths will change. What still holds should not have to start over.

of median provider latency
recomputations
provider/model families
blind third-party reconstruction
LimFlex Judgment Terrarium externalizes the structure from question to qualified judgment as a living state outside the AI that created it.

The model generates candidates. The qualified judgment lives outside the model.
Past judgments and judgments from other tools do not have to remain trapped inside old chats, old models or old workflows.

A better model may generate a better candidate. A different provider may choose a different reasoning path. None of that, by itself, should silently overwrite what is already qualified.
Keep what still holds. Review what is affected. Requalify what changed. Preserve unresolved conflict instead of forcing it into certainty.
