When to reach for AI generation

AI generation is best for routine assignments where you know the topic and the depth, but don't want to type out every bullet point. Capstone projects, research papers, and anything novel still benefit from being written by hand.

Writing a useful prompt

On the assignment editor, click Generate with AI. The modal asks for three things:

Topic — the subject of the assignment, e.g. “Database normalization through 3NF”.
Format — written report, lab exercise, problem set, presentation, or a custom value.
Depth — short answer, standard, or in-depth. This mostly controls how many bullets the model produces.

The model returns a markdown bullet list. Athenium parses lines that start with - and stores each as a separate requirement — so the editor renders them as a checklist students can tick off.

Steer with the topic field

The topic prompt is short but high-leverage. Adding a few words like “include at least one diagram” or “cite course material” routes through to the model and shows up in the generated bullets.

Reviewing the output

The first response is rarely perfect. Common adjustments:

Tighten verbs. Models love “explore” and “discuss” — replace them with verbs that imply a concrete deliverable (build, calculate, prove, compare).
Cap bullets at five. If the model returned eight, delete the weakest three. Long lists drag down completion rates.
Add a citation requirement. If you want sources, state it explicitly — the model won't infer.

Setting deadline and marks

Below the requirements, set a deadline and a max marks value (default 25). Athenium uses the deadline to drive notifications and to label submissions as on-time or late.

Publishing

Click Publish. Every member of the classroom gets a notification, plus an email if your admin has enabled Resend. Students see the assignment with a countdown to the deadline.

Editing after publish

You can edit requirements after publishing, but students who've already started won't see your changes highlighted. For non-trivial scope changes, post an announcement and consider extending the deadline.

Behind the scenes

Generation runs server-side through the Gemini API. Athenium sends the topic, format, and depth wrapped in a system prompt that constrains the output to a markdown bullet list. The response is parsed (lines starting with - ) and persisted to the assignment's requirements field as a string array.