Private Pro category · 1 compositions
Artificial Intelligence
AI-product explanation with honest capability and limit framing. These records are manifest-derived and source-free; an implemented status does not imply manual verification, release readiness, entitlement protection, or availability for purchase.
Manifest items
All 1 records currently carry implemented status. A dedicated detail becomes internally discoverable only after release readiness and a reviewed public preview.
Category contract
- Purpose
- AI-product explanation with honest capability and limit framing.
- Declared dependencies
- gummy-typography, gummy-button, gummy-card
- Required review
- light, dark, responsive, keyboard-reviewed, rtl-reviewed, reduced-motion
How to evaluate Artificial Intelligence blocks
Use this composition when an AI product needs to explain what the system does, what evidence supports it, and where human judgment remains necessary. It is a framing structure, not permission to promise accuracy, autonomy, safety, or outcomes the product has not demonstrated.
Connect capability statements to real workflows, inputs, outputs, and controls. Replace generic example copy with product-specific limitations, data-handling context, and a route to fuller documentation. Preserve the reading order when feature cards move from a wide grid to a single column.
Ask product, legal, privacy, and safety owners to review every claim before publication. Exercise long model names, translated copy, error states, keyboard navigation, reduced motion, and high zoom. The public preview remains unavailable until its separate review and release gates pass.
A useful handoff names the model or service version behind each example, the evaluation source for every measured result, and the owner of limitations and incident language. Include approved failure copy and escalation routes alongside the happy path. That record gives future reviewers a way to distinguish a product change from a purely visual update.
Boundary
Public pages contain only approved names, purposes, dependency aliases, requirements, status, and any future reviewed image preview. Editable paid source, test paths, release locations, and entitlement details remain private.