How much does the conclusion depend on the model?
The observatory recomputes the leading candidate set across declared ranking weights, climate boundaries, composition relations, atmospheric assumptions and direct-imaging trade cases.
25 candidates · five legacy weight menus · three climate prescriptions
25
candidates tested
same evidence, varied assumptions
1611
largest rank span
places across weight menus
3
climate prescriptions
agreement remains visible
3+
composition models
only where domains support them
What the analysis found
Some leaders are stable; some are choices in disguise
A ranking can look precise while depending strongly on its weights. The candidate-level view exposes the full rank interval and the ranges produced by supported physical models. Missing model support is shown as unsupported rather than converted into agreement.
Interactive robustness view
How much does the answer move?
A short bar can also mean the relevant model is unsupported; the text value carries that distinction. The ranges cover the implemented model menu, not every plausible model.
