Skip to main content
Eva is the creation assistant inside Atlas. Eva translates natural-language intent into a Protocol or Blueprint draft. Eva does not measure. It operates in the creation layer only: reading intent, surfacing nuance candidates, and proposing configurations for human review.

​
What Eva does

Protocol drafting. Describe what you want to measure. Eva returns a labelled, weighted nuance set with C×B×I tags and a zero-shot fidelity estimate. Nuance refinement. Edit labels or descriptions. Eva adjusts the spectral configuration to match the new intent without requiring a full redraft. Fidelity diagnosis. Eva flags nuances where the initial draft shows low fidelity and suggests exemplars or contrastive pairs to strengthen them. Blueprint assembly. Select Protocols from the library or draft new ones. Eva checks for nuance overlap and interaction coverage gaps.

​
Limits

Eva is not the measurement. Eva generates candidate configurations. Every Protocol requires a human publish decision before it becomes callable from the API.