← SkillSafe / Battle Simulator

Two contenders, a setting that actually matters, and numbers that came from somewhere.

Ask who wins between a grizzly and a silverback and almost any answer you get will be built on invented bite forces and confident body masses that are wrong in ways you cannot check. This one ships a stat table — 46 contenders, 205 published figures, each labelled for where it came from — and passes it into the request as fixed input. The reasoning is the model's. The arithmetic is not. Afterwards every number in the verdict is checked back against the table, and the ones that were not supplied are shown to you rather than passing as fact.

Each example ships with a saved verdict, so you can read a whole adjudication and its figure audit without signing in and without spending a credit.

Anything not in the table still runs — it is marked as untabled, the verdict is told to reason about it qualitatively and to state no figures for it, and any number that appears for it anyway is flagged in the audit.

The setting

These are not flavour text. Each axis is resolved against each contender's traits before the request is made, into a list of capabilities this setting amplifies, nullifies or pulls both ways — and that list is what the verdict reasons from. Change the terrain and the input genuinely changes.

Add a note about the setting (optional)

The stat table

Every figure the adjudication is allowed to state, and where each one came from. Read it before you trust a verdict built on it — a curated table is a claim about the world and this one is making it in the open.