Re-check with a fresh report after a detector update

Charles Li · HumanPen · recheck-after-detector-update

Detectors and portals ship model or legend updates. The literacy move here is re-check with a fresh report after a detector update before spending more rewrite rounds on an old map.

Distinct from reimport-report-after-pass (after your own edit), version-mismatch-ai-score (attempt identity), and cross-detector-report-one-pass (one import workflow across tools). Here the point is vendor change → new evidence.

Better order

1. Note when the portal or detector announces a model, legend, or coverage change.
2. Export a fresh AI writing report for the same attempt (or a new attempt if the school requires it).
3. Import that report; confirm attempt identity and read the new legend once.
4. Rewrite only continuous-prose spans flagged on the new report; freeze citations, numbers, formulas, and layout.
5. Do not keep editing from highlights that belong to a retired detector version.

Why old highlights go stale

A model update can re-paint which spans look flagged. Working from a pre-update PDF can send free rounds at the wrong sentences and confuse supervisors comparing screenshots.

What this is not

• Not a guarantee the next AI percentage will drop or clear.
• Not advice to shop detectors until one looks softer.
• Not a substitute for your course integrity process if one is opened.

Product: humanpen.net. Charles Li · HumanPen.