The ACADA Assistant

The ACADA Assistant is an AI that answers questions about how ACADA works and, inside a course, about the lesson you have open. Open it from the sparkle in the header, or with Alt+A. It runs on Anthropic's API, and it can be wrong.

What it knows

• How the platform works: the roles, the prices and shares, where things are. It links the help article when one answers the question.
• The course, module and lesson you have open, if you can open that course. It reads the lesson's text, not the answers to its quizzes.
• For creators, the Editor and the block types, so it can advise on building a course.

What it will not do

• Write work you will hand in. It is set up to explain, give an example on a different case, or comment on your own draft: see Academic integrity.
• Change anything. It cannot click, buy, post or edit for you.
• Give professional advice for your situation. On medical, legal, financial or safety questions it explains the subject and says so.

How to use it

1. 1.

   Open the panel and type in Ask the assistant…, or choose one of the suggested questions.

2. 2.

   To ask about a passage, select it in the lesson and choose Ask the Assistant.

3. 3.

   Keep an answer with Save to Notebook, or Copy it. New conversation starts afresh.

The panel docks at the side; Pop out into a window floats it, and Dock back to sidebar returns it. Alt+A closes it again.

What you can change

To hide the Assistant, turn off Show the ACADA Assistant in Settings → Appearance. The sparkle in the header, Alt+A and Ask the Assistant on selected text all go until you turn it back on there. The setting is kept in this browser.

Rules and limits

• You can send 8 messages a minute, 30 in any 5 minutes and 60 a day. The day starts again at midnight UTC.
• A message can be up to 8,000 characters. In a long conversation, the oldest turns drop out of what it sees.
• We keep a count of your messages, not the conversation. Nothing you ask trains an AI model.
• It needs you to be signed in.

If something goes wrong

"That is 8 messages in a minute…"

Wait a few seconds, then send again.

"You are sending messages faster than the assistant can answer. Give it a moment."

That is 30 messages in 5 minutes. Wait a few minutes, then send again.

"You have sent today's 60 messages to the assistant…"

The count starts again at midnight UTC.

"This conversation is too long. Start a new one."

Choose New conversation.

"The AI model's safety checks declined this question…"

Ask it another way. The checks sometimes decline harmless questions.