Constructor Prism: September 24, 2026

Overview

In this release, we added the following features:

• Activity descriptions that learners read before they start
• Word and PowerPoint files as sources for AI course generation

We also fixed issues affecting AI course generation, the course editor, the Content Library, and course settings.

What's new

A description learners read before they start

Authors can now write a description that appears above the activity card, so learners know what an activity is before they start an attempt. On video activities, the description appears above the player.

The description uses the same editor as media blocks and supports the full block set. It is available for labs, assignments, quizzes, surveys, SCORM packages, groups, and video. The description belongs to the activity in the Content Library, so every course that reuses the activity shows the same text. Description changes reach learners when the activity is published again.

In the Content Library, Open now leads to the activity page, where the description is edited. The activity's own editor opens from there through Edit Activity on the card. Creating an activity still opens its editor directly.

The existing description field is now labeled Author note and is not shown to learners. Existing text is not moved into the new description; to show it to learners, copy it into the description manually. Discussions and live events keep their existing description field.

Tracking IDs: LRN-21119, LRN-21203, LRN-21246

Word and PowerPoint files as course sources

The course generation chat now accepts DOCX and PPTX files as sources, alongside PDF, Markdown (MD), and plain-text (TXT) files and web links.

The file picker offers only the formats the agent can read. Files in other formats are rejected before upload with a message that names the format and lists the supported ones. For legacy DOC and PPT files, the message suggests saving them as DOCX or PPTX. Files can also be dropped onto the chat and are checked the same way.

Markdown and plain-text files are read as a whole and are limited to 1 MB. Larger material works best as PDF or DOCX.

Tracking IDs: LRN-21391, LRN-21392, LRN-21393, LRN-21394

Fixed issues

We have addressed the following issues:

AI course generation

• The agent now reflects edits that authors make to the body text of generated activities. Tracking IDs: LRN-21704
• When the AI usage limit is reached, the chat now shows a clear notice instead of a technical error. Tracking IDs: LRN-21026
• Course durations requested in minutes now appear correctly in the brief, and emoji now render on text slides of generated videos. Tracking IDs: LRN-21283, LRN-21267

Course editor and Content Library

• Publish changes is now disabled when all draft edits are undone, and Convert to course on a published composition now opens the new course's draft. Tracking IDs: LRN-17061, LRN-18123
• Duplicating a course or activity several times no longer gives every copy the same "2" in its name, and surveys now keep their thumbnail images. Tracking IDs: LRN-21668, LRN-21614
• Tooltips in Content Library activity preferences and the image upload progress on the Manage courses page now display correctly. Tracking IDs: LRN-19896, LRN-21951

Course settings and enrollments

• Numeric columns in the content flow are now right-aligned, and sections in enrollment settings can now be collapsed and expanded. Tracking IDs: LRN-17769, LRN-21733

Known issues

The AI authoring limitations described in the August 11, 2026 release notes still apply. The following issues remain open:

• A course published to Learn can show incorrect passing-score and score-weight values. Tracking IDs: LRN-21030
• A course can be published even if a quiz in that course has not yet been generated. Tracking IDs: LRN-21011
• The agent can generate videos of at least two minutes even when a shorter clip is requested. Tracking IDs: LRN-22043
• Markdown and plain-text source files larger than 1 MB are not read. The agent asks the author to split the file or attach it as PDF or DOCX.