# Stage 8: create and test your learning interface

**Outcome:** a tested app and honest publication decision. **Time:** 2 to 6
hours.

## Explain and inspect

Your interface is a view over governed knowledge, not a new authority. Use AI
as a coding collaborator while keeping the program small enough to understand.
Follow PRIMM: Predict, Run, Investigate, Modify and Make. Predict an Explorer
journey, run it, inspect the URL and data, then identify what your learner
needs less or more of.

## Do and check

Ask the AI to explain a minimal working interface. Modify one label, field and
interaction before making a view for your own user stories. Keep citations,
status and limits near the claims they qualify. Timelines, maps, comparisons,
guided trails, quizzes or source explorers may fit better than a chat box.

Test with somebody unfamiliar with the subject. Ask them to complete a task
without coaching and show where the answer came from. Check keyboard use,
focus, headings, contrast, narrow screens, missing data, broken links, slow
networks and unsupported questions.

Before public release, verify rights and privacy, freeze a version and digest,
run every gate, publish limitations, check the exact deployed URL and provide
a correction route. A local success is not a public release.

## Retrieve and reflect

Explain one decision made by you rather than the AI and its evidence. Record
what the bundle was good at and which other system remained authoritative.

Continue with the [beginner guide](../beginners/index.md) or the
[full review](../okf-evolution-review-2026-08-17.md).